User-Generated Platforms Glossary

Platforms where users supply the content or the inventory borrow their words from publishing, video engineering, payments, insurance, criminal law and regulation, and the same word shifts with context: a channel is a community on a streaming site and a room inside a server on a chat platform, a hold is a held chat message in one chapter and a card authorization or a legal preservation in others, and a label can be a content warning, an ad suitability tier or an AI disclosure. Entries are grouped by the part of the volume that explains them, alphabetized within each group, and each is one line and a link to the chapter that owns the idea; a word with two meanings appears once per meaning, qualified. The volume's own names for the core records (account, profile, community, content item, content revision, media asset, stream, listing, booking, moderation case) are the ones in The Core Data Model.

The Business#

• Account: one login, a person or a business, whatever roles it plays; everything else (mentions, follows, penalties, payouts) keys on its immutable ID, never on its handle (The User-Generated Platform Business).
• Account lock: moving an account to locked on a suspected takeover: sign-in refused, payouts held, the handle frozen and outbound messages stopped while recovery runs; a status, not a penalty, so the victim's standing stays clean (Accounts, Identity & Standing).
• Account standing: the user-facing summary of the penalties in force against an account (what it did, which rule it broke, what that restricts and when it expires), derived as a query over penalties rather than stored, and read at the moment of an action by the gates for monetization, badges, moderator roles and higher limits (Accounts, Identity & Standing).
• Account Standing tab: Discord's settings page, shipped in October 2023 with its Warning System, that lists an account's past violations and shows standing determined by active violations and their severity (Accounts, Identity & Standing).
• Account status: whether an account can be used at all, which is about control rather than conduct: active, locked (a takeover or recovery in progress), deactivated (paused by the user) or closed; every penalty, including a permanent ban, is a penalty row instead (Accounts, Identity & Standing).
• Account versus role: the split that one account reads, posts, moderates, streams, books and hosts, so the platform records which role it acted in for each act instead of creating separate accounts per role (The User-Generated Platform Business).
• Advertisers: brands and performance marketers buying attention next to user content, who need reach, targeting, measurement and suitability (The User-Generated Platform Business).
• Advertising: ads in feeds, conversations, search and video sold by auction; the main business of content platforms (about $2.1 billion of Reddit's $2.2 billion revenue in 2025) (The User-Generated Platform Business).
• API and data licensing: revenue from paid API tiers and multi-year content licenses sold to search and AI companies and developers (The User-Generated Platform Business).
• ARPU (average revenue per user): revenue divided by users; it varies by region by more than thirty times, as Pinterest's $30.84 in the US and Canada against $0.83 in the rest of the world shows (The User-Generated Platform Business).
• Badge: a visible mark on a profile; the rule is one visually distinct badge kind per meaning, each backed by a verification record, so revoking the verification closes the badge (Accounts, Identity & Standing).
• Bidirectional verification: the AT Protocol check that a handle resolves to a DID and that the DID document claims the handle; a handle failing either direction shows as handle.invalid (Accounts, Identity & Standing).
• Booking: a guest's reservation of a listing for a period, or a buyer's purchase; confirmed reservations of one listing can't overlap, which an exclusion constraint enforces in the database (The User-Generated Platform Business).
• Build, buy or rent: the per-layer vendor decision; the default is to build the community, content and marketplace model and the ranking, and rent video, image delivery, identity verification, payments and insurance (The Platform Landscape).
• Charge types: the connected-accounts decision a marketplace makes first, between destination charges, separate charges and transfers, and direct charges; it decides whose balance a refund or chargeback hits and whose name is on the card statement (The Platform Landscape).
• Classifier: a model that finds new, unjudged content and returns a score that should order a human review queue, unlike a hash match, which finds copies of something already judged (The Platform Landscape).
• Closure: the immediate part of account deletion: sign-in stops, the profile disappears, listings unpublish and the handle stops resolving, while erasure follows later as a job (Accounts, Identity & Standing).
• Commerce and affiliate: revenue from merchants through shoppable pins, product ads, catalog feeds and affiliate fees on attributed sales (The User-Generated Platform Business).
• Community: a space with members, rules and moderators (a forum, a server, a channel or a group), and the unit of scope for membership, rules, moderation and much of the load (The User-Generated Platform Business).
• Confusables: characters that look alike across scripts, such as Latin "a" and Cyrillic "а", mapped to one form before the handle uniqueness check so look-alike handles collide (Accounts, Identity & Standing).
• Connected accounts: a payments platform's product for paying many sub-merchants, which almost every marketplace runs on; Stripe Connect is the commonly documented example (The Platform Landscape).
• Content item: anything published that others can see, reply to or report (a post, a comment, a media post, a stream or a listing), carrying an author, a place, a visibility and a moderation state (The User-Generated Platform Business).
• Content revision: each version of an item's text as published, never edited; reports and evidence point to the revision they judged (The User-Generated Platform Business).
• Creator revenue share: a share of ad revenue on a creator's content, passed through to the creator; YouTube pays 55% of net watch-page ad revenue and 45% of a pooled Shorts allocation (The User-Generated Platform Business).
• Creators: people whose content draws the audience (posters, streamers, video makers, artists), who need reach, tools, a way to get paid and a fair hearing when moderated (The User-Generated Platform Business).
• Data licensees: search engines and AI companies paying for access to the corpus, who need reliable API access under contract (The User-Generated Platform Business).
• DAUq and WAUq (daily and weekly uniques): Reddit's audience measures: 121.4 million daily and 471.6 million weekly uniques in the fourth quarter of 2025 (The User-Generated Platform Business).
• Delegated verifier: an organization, such as a newsroom verifying its journalists, that grants badges on the platform's behalf; each badge records which verifier granted it, so a verifier's grants can be reviewed and revoked in bulk (Accounts, Identity & Standing).
• Deletion request: a user's request recorded with its scope (close the account, remove what they published, or erase their personal data), each fulfilled on its own clock (Accounts, Identity & Standing).
• Destination charge: a charge type where the platform takes the payment and transfers the host's share, which Stripe recommends for marketplaces such as home rentals (The Platform Landscape).
• DID (decentralized identifier): a permanent account identifier on the AT Protocol whose document lists the account's current host and keys; the handle is a human-readable name layered on it (Accounts, Identity & Standing).
• did:plc: a DID method in which the user can hold rotation keys (on paper, if they like) and repoint the DID to a new host without the old host's help, which is how an account moves (Accounts, Identity & Standing).
• Direct charge: a charge type where the seller is the merchant and the platform takes an application fee, suited to software platforms more than marketplaces (The Platform Landscape).
• Domain handle: a handle that is a domain name, proved by a DNS TXT record at _atproto.<handle> or an HTTPS response at /.well-known/atproto-did and re-checked on a schedule; Bluesky treats it as a form of verification, and over 309,000 Bluesky accounts used one by the end of 2025 (Accounts, Identity & Standing).
• Erasure job: the background job that walks each record type for a closed account, skips anything under a legal hold or retention window, erases or de-identifies the rest and reports what it removed and kept, re-running when a hold is released (Accounts, Identity & Standing).
• Federated network: a platform shape where posts live on independent servers that interoperate, read as timelines across servers (Mastodon and the fediverse, Bluesky on the AT Protocol), mostly funded by donations, hosting services and venture capital (The Platform Landscape).
• GBV (gross booking value): the total value of bookings made on a rental marketplace; Airbnb reported $91.3 billion for 2025 (The User-Generated Platform Business).
• GIFCT hash-sharing database: hashes of terrorist and violent extremist content, labeled by entity, behavior and incident and shared among members with an information-sharing agreement (The Platform Landscape).
• GMS (gross merchandise sales): Etsy's measure of the total value of goods sold on its marketplace: $10.46 billion in 2025 (The User-Generated Platform Business).
• GMV (gross merchandise value): the total value of goods sold on a goods marketplace; eBay reported $79.6 billion for 2025 (The User-Generated Platform Business).
• Guests and buyers: the demand side of a marketplace, who need an accurate listing, an all-in price and a refund when it isn't as described (The User-Generated Platform Business).
• Handle: a public, mentionable and impersonable name for an account; it's displayed as typed, normalized for uniqueness, and never used as the key other records point at (Accounts, Identity & Standing).
• Handle history: every handle an account has held, kept as time ranges so old names still redirect and resolve, with a released name held back from reuse for a period (Accounts, Identity & Standing).
• Handle key: the normalized form of a handle (case folded, confusables mapped) that carries the unique constraint, beside the handle as the user typed it (Accounts, Identity & Standing).
• Hosts and sellers: people and businesses that list inventory (a room, an RV, a car, a handmade item, a used jacket) and need bookings or sales, payouts on time and protection when something goes wrong (The User-Generated Platform Business).
• Identify-then-challenge: Airbnb's sign-in flow in which the user says who they are first and a server-side policy engine picks the challenge most likely to succeed, with a "try another way" fallback; it reported 2.6% more successful authentications and 27% fewer duplicate accounts (Accounts, Identity & Standing).
• Identity verified: the verification meaning in which a government ID, a selfie or third-party data was checked, shown to counterparties in a transaction (Accounts, Identity & Standing).
• Legal hold: a record that suspends erasure of named records for a stated reason (a claim window, an open dispute, a CyberTipline preservation, a preservation request, litigation, regulatory retention or tax) until it's released; one from a law enforcement request that forbids notice is never shown to the user (Accounts, Identity & Standing).
• Life-cycle events: events such as content.published, content.removed, booking.confirmed and account.closed, emitted through the outbox for feeds, search, notifications, payouts and transparency counts to subscribe to (The User-Generated Platform Business).
• Listing: the bookable or purchasable inventory side of a listing content item; whether it's visible is the content item's state, whether it can be booked is the listing's status (The User-Generated Platform Business).
• Live streaming platform: a platform shape where users supply broadcasts and others consume live video, chat and clips, earning from advertising, subscriptions, tips and virtual currency (Twitch, YouTube Live, Kick) (The Platform Landscape).
• Marketplace take rate: service fees charged to hosts, sellers, guests or buyers as a percentage of the booking or sale; Airbnb's revenue was about 13.4% of gross booking value in 2025 (The User-Generated Platform Business).
• Media asset: an uploaded file with its own processing state, hashes, storage cost and retention, attached to content items; one file can sit in several posts and be kept or purged on a different schedule from them (The User-Generated Platform Business).
• Media host: a platform shape where users supply images and video that others consume as galleries, embeds and feeds, earning from advertising and storage subscriptions (Imgur, Pinterest, YouTube, Vimeo) (The Platform Landscape).
• Membership: an account's membership in a community over time, with each join, leave and rejoin a separate range, so whether someone was a member when they posted has an answer (The User-Generated Platform Business).
• Moderation case: the playbook's case, report, decision and penalty tables, scoped to a community or the platform and pinned to the content revision that was judged (The User-Generated Platform Business).
• Paid checkmark: a badge granted for a subscription rather than a check; the European Commission fined X €120 million on December 5, 2025, in its first DSA non-compliance decision, finding that X's paid blue checkmark deceived users (Accounts, Identity & Standing).
• Paid communities: monthly tiers members pay for access to a server, channel or subforum; Discord's Server Subscriptions launched with owners keeping 90% (The User-Generated Platform Business).
• Peer-to-peer marketplace: a platform shape where users list stays, vehicles and goods that others book or buy, earning from take rate, seller ads, protection plans and float (Airbnb, Outdoorsy, RVshare, Turo, Etsy, eBay, Vinted) (The Platform Landscape).
• Penalty evasion: a banned user returning on a new account, detected by linking accounts to one person (devices, payment instruments, identity documents, behavior) (Accounts, Identity & Standing).
• Platform shapes: the six kinds of user-generated platform by what users supply and others consume: communities, media hosts, live streaming, chat, federated networks and peer-to-peer marketplaces (The Platform Landscape).
• Premium subscriptions: ad-free tiers, perks and badges that users pay for directly (The User-Generated Platform Business).
• Profile: a public face of an account (a member, a creator, a host, a seller or a brand), at most one of each kind per account, with its own role-specific state (The User-Generated Platform Business).
• Promoted listings and seller ads: paid placement inside a marketplace's own search, sold to sellers; eBay's $1,993 million of 2025 advertising revenue was "primarily" fees to promote listings (The User-Generated Platform Business).
• Protection margins: revenue from protection plans priced per trip, host plans that trade take rate for deductible, and deposit waivers (The User-Generated Platform Business).
• Protection plan: contractual protection a rental marketplace sells beside third-party insurance, such as Turo's host plans, which are contractual reimbursement and "not insurance" (The Platform Landscape).
• Re-verification: checking identity again after a change of legal name or other account information, or on detected risk; Airbnb blocks new reservations, or the host's calendar, until it's done (Accounts, Identity & Standing).
• Regulators and law enforcement: the FTC, state attorneys general, the European Commission and national coordinators, Ofcom, eSafety, the ICO, NCMEC as the US reporting channel, police and courts, who need reports in their format, notices answered on their clock and records on demand (The User-Generated Platform Business).
• Remote account: an account from another federated server that a server keeps rows for and moderates locally (limit, suspend) but can't delete at its home (Accounts, Identity & Standing).
• Reserved handle: a name held back from registration: brand and public-figure names, staff-like names such as "admin" or "support", names of banned accounts and recently released handles (Accounts, Identity & Standing).
• Role: a named set of community permissions with a rank, held through memberships; platform staff use separate grants, never community roles (The User-Generated Platform Business).
• Role-scoped penalty: a penalty that names the role it restricts (host, guest, member, moderator and so on), so a host who cancels repeatedly loses instant book or visibility rather than the account; sitewide suspension is reserved for conduct that makes the person unsafe in every role (Accounts, Identity & Standing).
• Seller subscriptions: monthly plans with tools and credits sold to sellers, such as Etsy Plus at $10 a month (The User-Generated Platform Business).
• Separate charges and transfers: a charge type where the platform charges first and decides later whom to pay, for multi-party or not-yet-assigned orders (The Platform Landscape).
• Step-up: requiring a recent strong factor before actions where value or identity moves: changing the payout account, email, phone or handle, adding a co-host or moderator, transferring a community's ownership, or starting a stream with a new key (Accounts, Identity & Standing).
• Stream: a live broadcast's state (scheduled, live, ended or interrupted) and its recording, kept beside the stream's content item (The User-Generated Platform Business).
• Strike: on YouTube, a violation after the first warning; each strike lasts 90 days, the first freezes posting for a week, the second for two weeks, and a third within 90 days can remove the channel (Accounts, Identity & Standing).
• Tipping and paid messages: one-time fan payments such as Super Chat, Super Thanks, Bits and tips; YouTube passes 70% of fan funding to the creator (The User-Generated Platform Business).
• Trusted Verifier: an organization Bluesky authorizes to verify accounts, whose badges are scalloped and name the organization that granted them; 21 Trusted Verifiers had verified 777 accounts by the end of 2025 (Accounts, Identity & Standing).
• User-generated platform: a platform where users supply what others consume (posts, media, streams or listings) and the platform makes those contributions findable, safe enough to use and worth paying for, while taking a cut, selling ads, or both (The User-Generated Platform Business).
• Users: people who read, watch, buy, book and occasionally post, often logged out, who need something worth their time, safety and control over their data (The User-Generated Platform Business).
• Vendors: video, CDN, identity, insurance, payments and trust and safety providers that integrate with the platform and receive its data under their terms (The User-Generated Platform Business).
• Verification: a record of what was checked about an account, by whom (the platform, a vendor or a delegated organization), with what evidence and for what period; "verified" can mean identity checked, authentic and notable, controls a domain, or paid (Accounts, Identity & Standing).
• Virtual goods and currencies: awards, boosts, cheers and gifts users buy with a platform currency; no platform in the survey breaks out the revenue (The User-Generated Platform Business).
• Volunteer moderators: users who run communities (forum moderators, server owners, channel mods) and need tools, automation and a platform that backs their decisions (The User-Generated Platform Business).
• Volunteer versus platform moderation: the split that on community platforms most removals are made by users, not staff; Reddit's report for the second half of 2025 attributes 52.7% of removals to moderators and 44.7% to admins (The User-Generated Platform Business).
• Warning: on YouTube, what a first violation usually gets; it expires after 90 days if the creator completes a policy training, and later violations become strikes (Accounts, Identity & Standing).
• WebFinger: the lookup that resolves a Mastodon user@domain handle: the server asks the domain's /.well-known/webfinger endpoint for the acct: URI, follows the self link to the ActivityPub actor, then looks up again from the actor's own username and host to confirm they match (Accounts, Identity & Standing).

Communities and Conversation#

• 2023 blackout: the June 12, 2023 protest in which thousands of Reddit communities (over 7,800 by one count) went private or restricted posting over Data API pricing, showing that moderators control availability and their tools are infrastructure (Volunteer Moderation & Community Governance).
• Abandonment: a community whose owner and moderators have stopped moderating and is drifting into spam; the platform hands it to an active member (checked against the owner's recent activity, with notice to the owner) or closes it (Communities, Roles & Permissions).
• ActivityPub: the W3C Recommendation (since January 23, 2018) behind Mastodon, the wider fediverse and Threads' opt-in federation, in which servers deliver activities such as Create, Follow and Delete to each other's inboxes by HTTP POST (Federation & Open Protocols).
• Actor: in ActivityPub, a user, group or service with an inbox and an outbox; posting adds an activity to the outbox and federation delivers it to recipients' inboxes (Federation & Open Protocols).
• Adjacency list: thread storage where each comment points at its parent and a subtree is read with a recursive query; paired with a materialized path, it's the default here (Posts, Comments & Voting).
• ADMINISTRATOR: the Discord permission bit that grants every permission and skips channel overwrites entirely (Communities, Roles & Permissions).
• AI answer: an answer a model composes from retrieved community threads, shown only with citations to the items it drew on and held to search's filters, so a removal of any cited item invalidates every cached answer citing it (Search & Discovery).
• Answer-ness: a search ranking signal that favors, for question-shaped queries, threads with accepted or highly voted answers over threads with only the question (Search & Discovery).
• API pricing: charging developers for API access; Reddit's terms effective July 1, 2023 kept free limits per OAuth client and charged $0.24 per 1,000 calls above them, with moderation tools exempted (Developer Platforms, APIs & Data Access).
• AppView: an AT Protocol service that indexes the network from the firehose and serves the aggregated views apps need: threads, like and repost counts, follower lists, search (Federation & Open Protocols).
• Archiving: closing old threads to new comments and votes after a period, which caps the surface spammers and manipulators can reach and lets scores freeze (Posts, Comments & Voting).
• AT Protocol: Bluesky's protocol, which separates where a user's data lives (a PDS) from the relays, AppViews, feed generators and labelers that index, rank and moderate it (Federation & Open Protocols).
• Automation share: the share of moderator actions made by automated tools; 68.6% of Reddit moderators' removals in the second half of 2025 were automated (Volunteer Moderation & Community Governance).
• Automod: a rules engine that volunteer moderators configure to remove, hold or report posts and comments matching conditions on the item and its author's account; it should offer structured rules first, linear-time regular expressions second, hard limits, a report-only mode and attribution of every action to its rule (Volunteer Moderation & Community Governance).
• AutoMod (Discord): Discord's structured automod with published limits per server: six keyword rules (up to 1,000 keywords and 10 regular expressions each) and one each for spam, keyword presets, mention spam and member profiles, with block, alert, timeout and block-interactions actions (Volunteer Moderation & Community Governance).
• AutoModerator: Reddit's per-community rules engine, configured by moderators, that removes, holds or reports items matching conditions on the item and on its author's account, such as account age and reputation (Volunteer Moderation & Community Governance).
• Base permissions: the first step of the permission algorithm: the default role's permissions ORed with those of every role the member holds (Communities, Roles & Permissions).
• Block: a one-way filter that hides each account from the other everywhere (feeds, replies, search, mentions, recommendations, chat), enforced in every service in both directions (Feeds & Recommendations).
• Blocklist (federation): a list of domain blocks shared between fediverse admins as a CSV file; it cuts off bad servers fast but propagates errors as fast, so imports should record their source and be reviewed (Federation & Open Protocols).
• Bot sharding: splitting a bot's gateway connections across shards, which Discord requires from 2,500 servers, assigning each server by shard_id = (guild_id >> 22) % num_shards (Real-Time Chat).
• Bridgy Fed: a bridge between ActivityPub, the AT Protocol and the web, opt-in per user: bridging is off until the user enables it and off again when they block it (Federation & Open Protocols).
• Brigading: members of one community descending on another to vote or comment, answered with links that don't allow voting from the referral and tools that show moderators where a surge came from (Posts, Comments & Voting).
• Bucket: a time window (about ten days in Discord's design) that, combined with the channel id, forms a message partition key so no partition grows past a comfortable size (Real-Time Chat).
• Candidate generation: the first recommendation stage, in which several sources (the viewer's graph, similar users, embedding neighbors, related and trending items) each cheaply propose hundreds to thousands of items (Feeds & Recommendations).
• Category: a channel that holds other channels and carries permission overwrites they can inherit (Communities, Roles & Permissions).
• Category sync: a channel following its category's overwrites, so changing the category changes every synced channel; editing the child's own overwrites breaks the sync (Communities, Roles & Permissions).
• Channel: a conversation space inside a community with its own audience, kind (text, forum, announcement, voice, stage, category) and permission overwrites (Communities, Roles & Permissions).
• Channel (streaming): one person's community on a streaming platform: the broadcaster owns it, appoints moderators and VIPs, and everything else is chat around the stream (Communities, Roles & Permissions).
• Channel overwrite: per-channel allow and deny bits for a role or a member, applied in a fixed order: the default role's (deny, then allow), the member's roles combined (all denies, then all allows), then the member's own (deny, then allow) (Communities, Roles & Permissions).
• Chat search: search over messages sharded by who searches: server messages indexed by server, DMs by user with each DM stored once per participant, results filtered by the searcher's current channel permissions (Real-Time Chat).
• Chronological graph feed: a feed of what the people and places a user chose posted, newest first, such as Bluesky's Following or a Discord channel (Feeds & Recommendations).
• Closure table: thread storage in a side table holding every ancestor-descendant pair, good for arbitrary subtree queries at the cost of a row per ancestor on every write (Posts, Comments & Voting).
• Code of conduct (moderators): rules the platform sets for volunteer moderators and enforces against them; Reddit's Moderator Code of Conduct actions include removing moderators, restricting a community and limiting posting, with 878 reports actioned in the second half of 2025 (Volunteer Moderation & Community Governance).
• Cold start: serving users, items or communities with no history: interests asked at sign-up and popular-in-region candidates for new users, reserved exploration and a proving ground for new items, directories and onboarding picks for new communities (Feeds & Recommendations).
• Community ban: a penalty scoped to one community with community_id set, appealed to the community's moderators through modmail first, not an account penalty (Volunteer Moderation & Community Governance).
• Community rules: rules a community adds on top of sitewide rules, published as versioned text and cited in removal notices; they can be stricter than sitewide policy, never looser (Volunteer Moderation & Community Governance).
• Community-level enforcement: the platform quarantining, restricting or banning a whole community that won't enforce sitewide rules; Reddit banned 605,570 communities in the second half of 2025, 76.7% of them for spam (Volunteer Moderation & Community Governance).
• Community-ranked feed: a feed ranked by a community's votes, such as a Reddit community's hot list (Feeds & Recommendations).
• Confidence: Reddit's "best" sort: the lower bound of the Wilson score interval for the share of upvotes, which rewards comments whose quality is likely given how few votes they have, and ignores age (Posts, Comments & Voting).
• Content rating: the platform's gating summary of an item or community (general, mature, adult); a community's rating is a floor, and an item can be rated higher than its community but never lower (Posts, Comments & Voting).
• Controversy: a sort that finds items that split the voters, scoring 500 up and 500 down at 1,000 and 1,000 up and 10 down at about 1.07; a reader's option, rarely a default (Posts, Comments & Voting).
• Crosspost: a post that republishes another item into a new place; like every repost it holds a reference to the original, never a copy (Posts, Comments & Voting).
• DAVE: Discord's protocol, announced September 17, 2024, that end-to-end encrypts audio and video in DMs, group DMs, voice channels and Go Live streams using a Messaging Layer Security group and per-sender media keys, while text stays under content moderation (Real-Time Chat).
• Defederation: suspending a whole domain on the fediverse, so nothing from it is stored except usernames and existing follows are removed (Federation & Open Protocols).
• Delete activity: the ActivityPub message a home server sends to every server that may hold a copy of a deleted post; receivers should remove it and may answer later requests with a Tombstone and 410 Gone, so erasure beyond your own servers is best effort (Federation & Open Protocols).
• Deleted: the content state for an author's own removal of their item, shown to readers as a placeholder saying the author deleted it (Posts, Comments & Voting).
• Deletion latency: the time from a removal decision to the item disappearing from search and cached answers, a safety property because a removed post that stays searchable is still published (Search & Discovery).
• Devvit: Reddit's developer platform for apps that run on Reddit; when it announced the end of public API access, it offered an App Migration Program with a $1 million fund, $1,000 payments to apps that move and free hosting (Developer Platforms, APIs & Data Access).
• DM settings per community: letting members refuse DMs from people whose only connection is a given community, closing the path from a large public server to a private conversation (Real-Time Chat).
• Domain block: a fediverse admin's action on a whole server: reject media (no files processed), limit (all its accounts limited, follows become requests) or suspend (defederation) (Federation & Open Protocols).
• Edit: a new content_revisions row that moves the item's pointer, while reports and decisions keep pointing at the revision they judged so an author can't edit a reported post into innocence (Posts, Comments & Voting).
• Embedding: a vector a model makes from an image region or whole image so that visual similarity becomes distance; embeddings are per model version, so a new model means re-embedding the corpus (Search & Discovery).
• Fan-out on read: building a following feed at read time by gathering recent posts from everyone the viewer follows, which makes writes cheap and reads expensive (Feeds & Recommendations).
• Fan-out on write: pushing each new post's id into every follower's precomputed feed list when it's published, which makes reads cheap and writes expensive in proportion to follower count; platforms skip it for high-follower authors (Feeds & Recommendations).
• Fediverse Enhancement Proposal (FEP): how ActivityPub implementations agree on features the specification leaves out, such as quote posts (Federation & Open Protocols).
• Feed definition: a feed (the platform's own or a third-party generator) stored as data, including whether it's profiled, so the non-profiled option is a row every surface can be pointed at (Feeds & Recommendations).
• Feed generator: an independent AT Protocol service that returns a ranked list of post identifiers for the AppView to hydrate, published as a record so anyone can build a feed and users subscribe to it like an account (Federation & Open Protocols).
• Firehose: the combined stream of updates a relay collects from many PDSes on the AT Protocol, which AppViews index (Federation & Open Protocols).
• Follow: a one-way edge needing no consent (to a person, community, tag or board), the primary input to following feeds and a strong recommendation signal (Feeds & Recommendations).
• Followers collection: an ActivityPub actor's followers, the default audience its activities are delivered to (Federation & Open Protocols).
• Forum: a community of posts, where membership is often a subscription more than a gate and content is meant to be found through indexing, ranking and recommendation (Communities, Roles & Permissions).
• Freeze: a Mastodon action that stops a local account from being used while keeping its content (Federation & Open Protocols).
• Freshness by intent: choosing a search result's time decay from the query's intent: hours-old posts for a news event, the best thread regardless of age for a how-to question (Search & Discovery).
• Friend-request warning: an alert on friend requests from unfamiliar senders; Discord's UK changes show one when there are no mutual friends or only a shared server of under 200 members (Real-Time Chat).
• Friendship: a mutual edge that one side requests and the other accepts, mattering more for who may message whom than for feeds (Feeds & Recommendations).
• Gateway: the one connection a chat client opens to receive everything for every place it belongs (messages, typing, presence, member changes, voice state, moderation actions), with heartbeats, intents and resume (Real-Time Chat).
• Group: a community shape between forums and servers, such as Facebook Groups: a feed of posts inside a membership gate, with admins, moderators and join questions (Communities, Roles & Permissions).
• Heavy ranking: the recommendation stage in which a large model scores each remaining candidate with many features for several predicted outcomes, such as click, long view, like, save, reply, share, hide and report (Feeds & Recommendations).
• Hidden: a temporary content state: collapsed by a spam filter, blurred behind a label, or held while a report is reviewed (Posts, Comments & Voting).
• Hot: Reddit's published sort ranking by the logarithm of net score plus age in units of 45,000 seconds (12.5 hours), so a post needs ten times the score to hold its rank against one 12.5 hours newer (Posts, Comments & Voting).
• HTTP Signatures: the de facto authentication of ActivityPub server-to-server requests: the sender signs with the actor's key and the receiver fetches the key from the actor document; Mastodon supports RFC 9421 HTTP Message Signatures by default since version 4.5.0 (Federation & Open Protocols).
• Impression log: the record of every served feed page with each item's candidate source, model version, position and demotions, which the next model trains on and which answers what was recommended to whom (Feeds & Recommendations).
• Inbox: the endpoint where an ActivityPub actor receives activities delivered by other servers (Federation & Open Protocols).
• Index from events: feeding the search index from content and visibility life-cycle events rather than polling, with removals on a separate higher-priority path with its own latency target (Search & Discovery).
• Installation: an app's presence in one community, granted permission bits that the installer must hold themselves and that moderators can narrow or end, stored as a range (Developer Platforms, APIs & Data Access).
• Instance: a community that is also a host, such as a Mastodon server, where the community and the operator are the same party with the duties of both (Communities, Roles & Permissions).
• Intents: the event groups a gateway client or bot declares it wants, as a bitfield, which cut fan-out at the source (Real-Time Chat).
• Interactions: slash commands and context menus that deliver exactly the input a user chose to send an app, so most apps never need to read every message (Developer Platforms, APIs & Data Access).
• Invite link: a link that lets a specific person or anyone holding it join, with an expiry and a maximum use count; recording which link each member used lets a leaked one be revoked and its users found (Communities, Roles & Permissions).
• Join gate: a condition checked at the moment of joining a community (a verified email or phone, a minimum account age, rules accepted, a paid tier or an age check), stored as data with its version (Communities, Roles & Permissions).
• Join policy: how people get into a community: open, by request with moderator approval, by invite, or gated on a condition (Communities, Roles & Permissions).
• Karma: an account's accumulated score from votes, used as a gate for posting and sometimes as a payout input; it should be derived from votes that survived manipulation checks, exclude self-votes, and never be the only gate (Posts, Comments & Voting).
• Label: a statement of what content contains so it can be blurred, hidden, kept from minors or kept from ads, kept as a row with its source: the author, moderators, the platform or a third-party labeler (Posts, Comments & Voting).
• Labeler: an independent AT Protocol service that attaches signed labels to accounts and records, which users subscribe to as they would follow an account (Federation & Open Protocols).
• Lexicon: a shared AT Protocol schema (com.atproto.* for repository sync, app.bsky.* for the social app) that lets independent services agree on record and API shapes (Federation & Open Protocols).
• Light ranking: the recommendation stage in which a cheaper model trims thousands of candidates to hundreds before heavy ranking (Feeds & Recommendations).
• Limit: a Mastodon action that shows an account only to its followers on that server; as a domain block, it limits every account from a server and turns follows into requests (Federation & Open Protocols).
• Locked thread: a thread a moderator has closed to new replies, a control layered on top of content states (Posts, Comments & Voting).
• Logged-out reader: a visitor who never signs in, usually from search, served related items computed per page and cached, with personalization only under consent; Reddit monetizes them with ads on conversation pages and in feeds (Feeds & Recommendations).
• Mastodon account actions: a server's local actions on an account: mark media sensitive, freeze, limit or suspend (data kept 30 days before deletion), emailed to the user with one appeal within 20 days (Federation & Open Protocols).
• Materialized path: thread storage in which each row stores its ancestry as a string or ltree, so a subtree is one ordered index range scan (Posts, Comments & Voting).
• Member list (visible range): sending a client only the slice of a large community's member list on screen rather than presence for every member (Real-Time Chat).
• Mesh: a real-time topology in which every participant sends to every other, which works for three or four people and not beyond (Voice & Live Rooms).
• Message content intent: Discord's privileged intent that delivers message text to an app; without it, content fields arrive empty except in messages the app sent, DMs with it, messages that mention it and messages a context-menu command was used on (Developer Platforms, APIs & Data Access).
• Message request: a separate inbox for DMs from strangers, opened only when the recipient accepts; Discord routes DMs from non-friends who share only servers of 200 or more members there, and ignoring a request never notifies the sender (Real-Time Chat).
• Mixer (MCU): a real-time server that decodes every participant, mixes audio and tiles video, and sends each one combined stream: cheap for clients, expensive for the server, and slower (Voice & Live Rooms).
• Mod log: the append-only record of every moderator action in a community, automated or manual, with the actor's rank and the rule and reason cited; the community's accountability record (Volunteer Moderation & Community Governance).
• Mod queue: the queue of reported and automod-filtered items for a community's moderators, the playbook's reports and queues with a community scope (Volunteer Moderation & Community Governance).
• Moderation recording: recording a room for safety review, kept briefly and listened to only on a report, with notice, short default retention and access logging; impossible server-side under end-to-end encryption (Voice & Live Rooms).
• Moderator concentration cap: a limit on how many large communities one person may moderate, held as configuration by community size band; Reddit introduced one in 2025 (Volunteer Moderation & Community Governance).
• Modmail: the conversation channel between members and a community's moderation team as a team, used for appeals, removal questions and reports, so a moderator's departure doesn't lose the conversation (Volunteer Moderation & Community Governance).
• Mute: a one-way filter that hides the target from the muter's own feeds without telling anyone (Feeds & Recommendations).
• Nested sets: thread storage with left and right numbers per node: fast reads, expensive inserts, and a poor fit for live threads (Posts, Comments & Voting).
• Non-profiled feed: a feed option not based on profiling, which DSA Article 38 requires very large platforms to offer for each recommender system; built as a first-class mode stored on the account and honored on every surface (Feeds & Recommendations).
• Notification budget: limits on recommendation-driven pushes: a per-user cap within a hard ceiling, quiet hours in the user's time zone, consent, and a re-check of eligibility at send time (Feeds & Recommendations).
• On-platform app runtime: hosting developers' code inside the platform rather than serving an external API, so data never leaves, at the cost of sandboxing, quotas, review and availability becoming the platform's job (Developer Platforms, APIs & Data Access).
• One-party consent: the US federal rule (18 U.S.C. 2511(2)(d)) that a party to a communication may record it; some states require all parties, so a room with participants from many states follows the strictest rule (Voice & Live Rooms).
• Outbox: the collection an ActivityPub actor adds activities to when it posts, likes, follows or deletes (Federation & Open Protocols).
• Ownership transfer: handing a community to a new owner, irreversible and treated like a payout account change: re-authentication, the recipient's confirmation, notice to moderators and a short window for the old owner to cancel (Communities, Roles & Permissions).
• PDS (Personal Data Server): the AT Protocol service that hosts a user's repository of signed records and manages their identity (Federation & Open Protocols).
• Pending review: the content state where a filter puts an item it isn't sure about (an automod rule, a new account's first post, a classifier's uncertain middle) until moderators approve or remove it (Posts, Comments & Voting).
• Permission algorithm: the deterministic evaluation of a member's permissions in a channel: the owner has everything; base permissions from roles; channel overwrites in fixed order; implicit rules; and timeouts leaving only reading (Communities, Roles & Permissions).
• Permission bits: permissions stored as bit flags; Discord defines flags from bit 0 to bit 52 and serializes the set as a string, so a signed 64-bit column will eventually run out (Communities, Roles & Permissions).
• Personalized recommendation feed: a feed ranked by a model of each viewer, from candidates far beyond their graph, such as Pinterest Home Feed, X's For You and YouTube Home (Feeds & Recommendations).
• PinSage: Pinterest's 2018 graph convolutional network over its pin-and-board graph that learns an embedding per pin, turning related-item retrieval into nearest-neighbor search (Feeds & Recommendations).
• Pixie: Pinterest's 2017 system of in-memory random walks over the pin-and-board graph, generating candidates without a trained model (Feeds & Recommendations).
• Platform staff access: a separate, audited, just-in-time grant that lets trust and safety, support and engineering staff act in any community without ever appearing as a role inside one (Communities, Roles & Permissions).
• Policy demotion: a ranking penalty applied after scoring to allowed but borderline items, recorded as its own reason so it's explainable, countable and testable (Feeds & Recommendations).
• Poll: a post format with options and a closing time, allowing one poll vote per account, kept apart from the item's up and down votes (Posts, Comments & Voting).
• Portability: moving an account between hosts: on the AT Protocol the DID is repointed and the account, posts and followers move together; on Mastodon only followers move, posts stay behind, and the old account becomes inaccessible (Federation & Open Protocols).
• Privileged intents: gateway event groups that need declaration and, at scale, approval: server presences, server members and message content; Discord requires verification for apps in 100 or more servers and approval of each privileged intent (Developer Platforms, APIs & Data Access).
• Proving ground: a "new" or "rising" pool seen by a small share of readers, from which posts that beat the section's baseline graduate to the main gallery; where vote-ring detection has to run first (Posts, Comments & Voting).
• Query log: a user's search history, personal data kept raw only briefly, then aggregated or de-identified, with safety-list matches counted separately and the history included in export and erasure (Search & Discovery).
• Rank: a role's position, deciding whom an actor can act on: assign only roles below their own highest, edit only lower roles, grant only permissions they hold, and kick, ban or rename only members below them; rank doesn't affect permission math (Communities, Roles & Permissions).
• Re-engagement notification: a push or email that recommends content rather than reporting something addressed to the user, chosen by the same ranking models as the feed and needing its own budget (Feeds & Recommendations).
• Read state: a member's last-read message per channel, updated on every scroll, batched per session and written behind, from which unread badges are derived (Real-Time Chat).
• Reblog: a repost that adds the reposter's own commentary on top of the chain before it, the mechanism Tumblr built its distribution on (Posts, Comments & Voting).
• Recording consent: notice before capture and a visible indicator throughout a recorded room, with a consent record per participant per session; a participant who declines listens without speaking, and apps that capture follow the same rule (Voice & Live Rooms).
• Relay: an AT Protocol service that collects updates from many PDSes into one combined stream, the firehose (Federation & Open Protocols).
• Remote actor: a user from another federated server, mirrored as a shadow accounts row with no login so posts, follows, reports and penalties reference one kind of account (Federation & Open Protocols).
• Removal reason: a reason a community defines (usually one per rule) that a moderator picks when removing and the author's notice states (Volunteer Moderation & Community Governance).
• Removed: the content state for a moderator's or the platform's removal; the author usually still sees the item and a notice of why, and everyone else sees a placeholder saying it was removed (Posts, Comments & Voting).
• Repost: a post pointing at another item; the original controls the copy, so when it's removed or deleted every repost and quote stops showing its content (Posts, Comments & Voting).
• Request coalescing: answering concurrent requests for the same rows with a single database query whose result they all share, Discord's defense against hot channels (Real-Time Chat).
• Researcher access: the DSA duty (Article 40) for very large platforms to give vetted researchers data on systemic risks on a coordinator's reasoned request, and access without undue delay to publicly accessible data; the Commission adopted the delegated act on July 2, 2025 (Developer Platforms, APIs & Data Access).
• Resume: reconnecting a dropped gateway session with its session id and last sequence number so the server replays what was missed instead of a full reload (Real-Time Chat).
• Retraction by negation: withdrawing an AT Protocol label by publishing the same source, subject and value with neg set, which keeps the history auditable (Federation & Open Protocols).
• Selective forwarding unit (SFU): a real-time server that receives each participant's stream and forwards the ones each receiver needs without decoding them, paired with simulcast or scalable video coding; community platforms use SFUs (Voice & Live Rooms).
• Self-label: a label the author sets on their own content; on the AT Protocol its authenticity comes from the repository's signature (Posts, Comments & Voting).
• Sensitive-media blur: blurring unsolicited images by default for teens and as a setting for adults; Discord's Teen Safety Assist turned it on in October 2023 (Real-Time Chat).
• Server: a community of channels, where membership is a gate and each channel decides who among the members can see it; most content isn't meant to be found by outsiders (Communities, Roles & Permissions).
• Server Subscriptions: Discord's paid community tiers giving members subscriber-only channels and roles, granted and revoked by the subscription's events, with eligibility rules for the owner (Communities, Roles & Permissions).
• Shared inbox: a server-wide ActivityPub inbox to which a sender delivers one copy per server for public and follower-addressed activities instead of one per follower (Federation & Open Protocols).
• Simulcast: a sender uploading several qualities of the same stream so an SFU can forward each receiver the one it can handle; scalable video coding does the same with one layered stream (Voice & Live Rooms).
• Slowmode: a per-channel minimum interval between a member's messages, from 0 to 21,600 seconds on Discord and also limiting thread creation, with a bypass permission for moderators (Real-Time Chat).
• Snowflake: Discord's 64-bit id whose top 42 bits are milliseconds since the platform's epoch, followed by worker, process and sequence bits, so ids sort by time and need no coordination to generate (Real-Time Chat).
• Sockpuppet: an alternate account used to vote for oneself (Posts, Comments & Voting).
• Stackable moderation: Bluesky's model in which users subscribe to moderation services, its own by default as a floor, and add independent labelers whose labels they configure to hide, blur or warn (Federation & Open Protocols).
• Stage: a room that separates a few speakers from an audience, with moderators inviting and removing speakers and listeners requesting to speak; Discord's stages allow up to 10,000 listeners (Voice & Live Rooms).
• Suspend: a Mastodon action removing an account from public view with its data kept 30 days before permanent deletion; as a domain block, it defederates a server (Federation & Open Protocols).
• Tags and hashtags: user-supplied facets normalized like handles, capped per item and moderated as content, because a tag can become a meeting point for prohibited content (Search & Discovery).
• Three searches: community search as three problems: content (posts and comments), places (communities and channels) and people (accounts and handles), each with its own ranking and safety rules (Search & Discovery).
• Timeout: a community penalty that strips every permission except viewing the channel and reading history, with owners and administrators exempt; stored as a feature restriction, so it expires with no clean-up (Communities, Roles & Permissions).
• Tombstone (comment): the placeholder a removed or deleted comment leaves in its thread, keeping its place so replies still have a parent, while its text is cleared from what's served (Posts, Comments & Voting).
• TransAct: Pinterest's 2023 transformer over a user's real-time sequence of recent actions, combined with batch-computed long-term embeddings, used in Homefeed, Related Pins, Notifications and Search (Feeds & Recommendations).
• Translation (machine): a derived rendering of a post in another language, labeled as machine-translated, invalidated when the original changes, kept out of evidence, and recorded with the engine and version that produced it (Search & Discovery).
• Two tiers of rules: sitewide rules that apply everywhere plus community rules on top; every removal records which tier and rule it enforced, and a community can be stricter but never looser (Volunteer Moderation & Community Governance).
• Typeahead: prefix search on community names and normalized handles, ranked by size and activity, filtered for visibility and blocks, and conservative about minors (Search & Discovery).
• Value model: the weights that combine a heavy ranker's predicted outcomes into one score, including negative weights on hides, reports and unfollows; the platform's editorial policy expressed as arithmetic (Feeds & Recommendations).
• Visual search: answering "more like this" and "what is this" for images through object detection, embeddings, approximate nearest-neighbor search and re-ranking with safety filters (Search & Discovery).
• Voice channel: a persistent room inside a community that members drop in and out of without a host; Discord's allow up to 99 users (Voice & Live Rooms).
• Vote: one account's judgment of one item (up, down or withdrawn), stored as a row so a later finding can discount it and recompute scores and karma; on Reddit every item starts with its author's upvote (Posts, Comments & Voting).
• Vote discount: zeroing a set of votes found to be manipulated and recomputing scores and karma, possible because votes are rows and scores are derived (Posts, Comments & Voting).
• Votes per view: votes divided by discounted views, a better quality signal for gallery ranking than raw votes (Posts, Comments & Voting).
• Workplace chat: chat such as Slack or Microsoft Teams that runs the community architecture with the duties inverted: the customer organization controls the data, moderates, sets retention and legal holds and answers legal requests (Real-Time Chat).

Media#

• ABR (adaptive bitrate) streaming: a player switching between a video's renditions as the viewer's bandwidth drops or recovers, which is why every video is encoded into a ladder of rungs (Video on Demand).
• Ad break: a cue in a live stream that inserts ads, recorded with its requester, requested duration, actual duration and the vendor's reference so creator ad revenue, advertiser billing and disputes reconcile against it (Live Latency, Playback & VODs).
• Alt text: a description of an image for screen readers, stored on the post's use of the file (content_media) rather than on the file, because one image can illustrate two posts that need different descriptions (Images at Scale).
• Anonymous upload: a file with no owner account (a null owner_account_id), which nobody can manage or be notified about, and so is the first candidate for a retention purge (Storage, Dedup & Delivery Economics).
• Archive tier: storage classes such as S3 Glacier Flexible Retrieval and Deep Archive that retrieve in minutes to hours, with 90 and 180 day minimums, for evidence and originals that are never served to a viewer (Storage, Dedup & Delivery Economics).
• Audio track: an alternate audio rendition of a video, such as a dubbed language or an audio-description track, carried in HLS as an EXT-X-MEDIA audio rendition with a source, a language and a state (Video on Demand).
• Automatic captions: captions generated by speech recognition, which YouTube publishes when ready and lets creators edit or remove; an automatic track should never overwrite one the creator uploaded (Video on Demand).
• AV1: a video codec published in 2018 by AOMedia under a royalty-free patent policy, about 30% more efficient than VP9 and HEVC by Meta's measurement, but expensive to encode and heavy to decode on older phones (Video on Demand).
• AVIF: an image format that gives the smallest files for photographs but is slow to encode, so it pays off for images viewed many times and often not for the long tail (Images at Scale).
• Blob: the stored object under one or more media assets: every asset points at a blob, exact duplicates share one, and it is purged only when no referencing asset is live and no legal hold covers any of them (Storage, Dedup & Delivery Economics).
• Blocking scan: a check the scanning gate waits for, such as hash-list matching, which fails closed: if the service is down, uploads wait in scanning rather than publishing and scanning later (The Upload Path).
• Byte-range playlist: an HLS playlist (EXT-X-BYTERANGE) that addresses each segment as a range of a larger file, so a rendition can be one object and a clip can be a playlist over a recording rather than a copy (Video on Demand).
• C2PA Content Credentials: a signed manifest recording an asset's provenance (origin, edits, AI use), bound to the content by a cryptographic hash so any alteration breaks the binding; validation says who signed it, not that the content is true (Images at Scale).
• Cache purge: a request to evict a file's cached copies from a CDN, every variant and format included; a takedown isn't done until a re-fetch confirms the cache no longer serves it (Storage, Dedup & Delivery Economics).
• Channel points: a channel-scoped points currency viewers earn by watching and spend on that channel's rewards, polls and predictions; run as an append-only ledger and kept non-monetary (never bought, transferred or cashed out) (Live Chat & the Interactive Layer).
• Channel type (IVS): Amazon IVS's per-channel processing choice (full transcode, transcode to a 720p or 480p ceiling, or transmux only), each priced per input hour, with prices differing about tenfold between the cheapest and dearest (Live Streaming: Ingest & Transcoding).
• Chat mode: a channel-wide gate the broadcaster or a moderator sets on live chat, such as slow mode, followers-only, subscribers- or members-only, emote-only, unique chat or approved users only (Live Chat & the Interactive Layer).
• Clear-key encryption: segment encryption with HLS EXT-X-KEY (AES-128 or SAMPLE-AES) and a key URL behind authentication, which stops casual downloading of segments (Video on Demand).
• Client-side multi-encode: having the creator's broadcast software encode the whole ladder on their own GPU and send each rendition as a track over Enhanced RTMP, so the platform transmuxes instead of transcoding (Twitch Enhanced Broadcasting, IVS multitrack video) (Live Streaming: Ingest & Transcoding).
• Clip: a short cut of a live stream made by a viewer or the streamer, published as its own content item with its own author, reports and life, pointing back to the source stream and able to outlive its recording (Live Latency, Playback & VODs).
• CMAF (Common Media Application Format): ISO/IEC 23000-19: one set of fragmented MP4 segments referenced by both an HLS playlist and a DASH manifest, so every device requests the same object under the same cache key (Video on Demand).
• Codec portfolio: the set of codecs (H.264, VP9, HEVC, AV1) a platform keeps so it can serve each device the best one it decodes; each codec is another ladder and more storage, so better codecs are gated on watch time (Video on Demand).
• Cold tier: millisecond-retrieval storage classes with a 90-day minimum, such as S3 Glacier Instant Retrieval, for long-tail originals that almost never load (Storage, Dedup & Delivery Economics).
• Common Encryption (CENC): ISO/IEC 23001-7, which defines four encryption schemes of which two are used: cenc (AES-CTR) and cbcs (AES-CBC with a subsample pattern); cbcs support across FairPlay, Widevine and PlayReady lets one CMAF set serve all three (Video on Demand).
• Content class: a classification of an image (photo, graphic, screenshot, artwork) recorded with the asset to tune quality, since screenshots and line art need a near-lossless path and photographs take aggressive perceptual quality (Images at Scale).
• Content disarm: publishing pixels the platform decoded and re-encoded instead of the uploaded bytes, which removes most of what a crafted file carries and strips metadata, though the decoder doing it still needs a sandbox (The Upload Path).
• Content-addressed storage: storing an object under the cryptographic hash of its bytes, so byte-identical uploads share one stored object; it saves nothing for re-encoded, resized or re-saved copies (Storage, Dedup & Delivery Economics).
• Convex hull: in per-title encoding, the upper-left boundary of a video's trial encodes plotted as quality against bitrate, where no other encode gives better quality for fewer bits; ladder rungs are picked from it (Video on Demand).
• Cost per thousand views: a surface's delivery cost per thousand views (roughly bytes served per view times the egress price, plus per-request and per-transformation fees, divided by how often the CDN hits), tracked per surface rather than as a monthly total (Storage, Dedup & Delivery Economics).
• Cost per viewer-hour: what it costs to encode, store and deliver an hour of watched video, the unit that ties the video pipeline to the business; delivery is usually the largest part at scale (Video on Demand).
• Creator export: a creator downloading their own uploads, which returns a rendition rather than the original if the platform discarded the source, and is refused for removed or struck content (Video on Demand).
• Crypto-shredding: deleting a file by destroying its own encryption key, which makes copies in erasure-coded volumes and backups unreadable without rewriting them, as Meta's f4 did per BLOB (Storage, Dedup & Delivery Economics).
• Cue: something that has to happen at a point in the video rather than at a wall-clock time (an ad break, timed metadata, a marker, a chapter, a stage change), recorded as a row for reconciliation (Live Latency, Playback & VODs).
• Custom reward: a channel-points reward a channel defines (Twitch allows up to 50), with a cost of at least 1 point, optional per-stream and per-viewer limits and a cooldown; a redemption stays unfulfilled until fulfilled or canceled, and canceling refunds the points (Live Chat & the Interactive Layer).
• DASH (MPEG-DASH): an adaptive streaming standard ratified by ISO in April 2012 and owned by no one company, with an XML manifest (the MPD), templated segment URLs, common encryption and multiple base URLs for multi-CDN delivery (Video on Demand).
• Decompression bomb: a small file that decodes into an enormous image, such as a uniform-color PNG declaring huge dimensions, guarded against by checking declared pixels against a cap before allocating anything (The Upload Path).
• Delivery restriction: a dated row the CDN layer reads to refuse or limit delivery (a geo-block, a referrer allowlist, signed-only access, an embed block), scoped to the platform, a community or an asset and carrying its legal basis (Storage, Dedup & Delivery Economics).
• Direct-to-storage upload: the pattern in which the API grants a scoped, expiring credential for one platform-named key and the client sends bytes straight to storage or a vendor, so the application server never carries them (The Upload Path).
• DRM (digital rights management): encryption whose keys are released only to a trusted decryption module on the device, through Widevine, PlayReady and FairPlay license servers; a cost and a constraint, used where a contract or paid content requires it (Video on Demand).
• Drops: watch-to-earn campaigns in which a game studio grants an in-game reward of no monetary value to viewers who watched participating channels (or subscribed) once their platform and game accounts are linked (Live Chat & the Interactive Layer).
• Dual format: sending a horizontal and a vertical layout of one broadcast at once over a multitrack connection, so desktop viewers see landscape and a phone held upright sees vertical (a Twitch beta from October 2025) (Live Streaming: Ingest & Transcoding).
• DVR (live rewind): letting viewers rewind within a stream while it is still live, set per stream; YouTube limits it on streams over 12 hours and Mux doesn't recommend it over four hours (Live Latency, Playback & VODs).
• Eager variant: an image variant made when the upload finishes, before the asset goes ready, giving predictable latency and storage for the sizes every new post needs (Images at Scale).
• Edge scanning: a CDN comparing content served through its cache against known child-safety lists, as Cloudflare's CSAM Scanning Tool does, blocking matches where it can and notifying the customer daily (Storage, Dedup & Delivery Economics).
• Egress: data transferred out to the internet, which most clouds bill (Amazon S3 after the first 100 GB a month) and Cloudflare R2 doesn't, so the provider's meter decides which behaviors cost money (Storage, Dedup & Delivery Economics).
• Encoding ladder: the set of renditions a video is encoded into, with a resolution, bitrate and frame rate per rung, kept as versioned data so a player can step down and up as bandwidth changes (Video on Demand).
• Enhanced RTMP: the Veovera extension of RTMP (v2 dated January 31, 2026) adding FourCC signaling for codecs such as HEVC, VP9 and AV1, HDR, multiple tracks in one connection and a server-initiated reconnect request (Live Streaming: Ingest & Transcoding).
• Erasure coding: storing data as coded fragments (Reed-Solomon within a datacenter, XOR across datacenters in Meta's f4) instead of full replicas, cutting the effective replication factor from 3.6 to 2.8 or 2.1 for warm data (Storage, Dedup & Delivery Economics).
• EventSub: Twitch's event subscription catalog (chat, holds, raids, rewards, polls, predictions, Hype Train, Shield Mode, ad breaks, drop grants and more), delivered at least once over webhooks, WebSockets or conduits with a message ID for de-duplication (Live Chat & the Interactive Layer).
• Eviction order: the order in which a retention cap deletes recordings or uploads, such as least viewed first (which protects what audiences value) or oldest first (simpler, and harder on long-time creators) (Live Latency, Playback & VODs).
• Exact dedup: deduplication keyed on a cryptographic hash of the original's bytes, done on the server after the full upload arrives so neither the response nor its timing reveals that someone else uploaded the same file (Storage, Dedup & Delivery Economics).
• EXIF Orientation: the metadata tag (values 1 to 8) recording how a phone's stored frame should be rotated or mirrored; the pipeline applies it to the pixels before stripping the tag, or the variant comes out sideways (Images at Scale).
• Extension: third-party web code in the viewing experience, shown as a panel below the video, an overlay over it or a component inside part of it, running in a sandboxed iframe on a separate origin (Live Chat & the Interactive Layer).
• FairPlay Streaming: Apple's DRM for HLS on Apple platforms, for which Apple approves production credentials only for teams providing a consumer streaming service (Video on Demand).
• Fanout: delivering one chat message to everyone in a room, published once to the channel's topic and written by each edge node to its local sockets, so cost and load scale with deliveries rather than sends (Live Chat & the Interactive Layer).
• Fast start: an MP4 with its moov index at the front, so the platform can probe duration, codecs and dimensions from the first bytes and begin work before the whole file is read (The Upload Path).
• Format negotiation: choosing an image format per viewer, either in markup (<picture> alternatives) or on the server from the request's Accept header, which then requires the cache to vary by it (Images at Scale).
• Geo-block: refusing delivery by the viewer's country at the edge; on a media host it blanks the platform's files on every embedding site in that country, and it is served as a 451 and recorded with its legal basis (Storage, Dedup & Delivery Economics).
• GIFV: Imgur's .gifv extension, introduced on October 9, 2014, for GIF uploads converted to WebM or MP4 video, the same move as serving animation as silent looping video at platform scale (Images at Scale).
• Glass to glass: live latency measured from the camera capturing a frame to a viewer seeing it, stated by platforms as what most viewers get because the viewer's network and player buffer are part of it (Live Latency, Playback & VODs).
• H.264 (AVC): the video codec that decodes on essentially every device, the universal fallback and the format of upload-time encodes, and the least efficient in the portfolio (Video on Demand).
• Hard binding: the part of a C2PA credential that binds the manifest to the content by a cryptographic hash, so a re-encoded variant no longer matches the original's binding (Images at Scale).
• Hate raid: many new or throwaway accounts arriving in a channel at once to flood its chat with abuse, answered with a preset lockdown and a mass action keyed on what the attackers did (Live Chat & the Interactive Layer).
• HEIC (HEIF): the high-efficiency photo format iPhones from the iPhone 7 capture when the camera is set to High Efficiency; only Safari renders it, so platforms accept it as input, keep the original and serve every variant in web formats (Images at Scale).
• Held message: a chat message kept invisible to the room by a term list, classifier or restriction until a moderator allows or denies it, recorded with the layer, model version and decision (Live Chat & the Interactive Layer).
• HEVC (H.265): Apple's high-efficiency and HDR video path, with its own ladder in Apple's HLS specification, allowed only in fragmented MP4 in HLS and not covered by a royalty-free policy like AV1's (Video on Demand).
• Highlight: a longer edit of a past broadcast made by the creator and kept after the past broadcast expires; Twitch caps highlights and uploads at 100 hours per channel since 2025 (Live Latency, Playback & VODs).
• HLS (HTTP Live Streaming): Apple's adaptive streaming protocol, published as RFC 8216 in August 2017, in which a multivariant playlist lists variant streams and each media playlist lists one rendition's segments (Video on Demand).
• Hold-back: the HLS rule (HOLD-BACK) that a player starts no closer to the live edge than at least three target durations, which with six-second segments puts standard HLS around 25 to 30 seconds behind once encoding and CDN time are added (Live Latency, Playback & VODs).
• Hot tier: storage classes with millisecond retrieval and no minimum duration (S3 Standard, R2 Standard), for new uploads, active variants and anything a page may request (Storage, Dedup & Delivery Economics).
• Hotlinking: another site's page loading a file directly from the platform's URL, which costs egress without an ad impression; platforms leave it open, limit it by referrer or require signed URLs, by content visibility (Storage, Dedup & Delivery Economics).
• HTTP 451: the status code RFC 7725 defines as "Unavailable For Legal Reasons," returned for a file blocked by a legal demand with a Link header (rel="blocked-by") naming the blocking entity (Storage, Dedup & Delivery Economics).
• Hype Train: a time-boxed group event on a channel in which contributions (cheers, subscriptions and gifted subscriptions) add points toward levels, expiring if contributions stop and followed by a cooldown (Live Chat & the Interactive Layer).
• I-frame playlist: an HLS playlist of I-frames only (EXT-X-I-FRAME-STREAM-INF) that players use for scrubbing and fast-forward, required by Apple's authoring specification (Video on Demand).
• IDR interval: the spacing of IDR (keyframe) frames, set to two seconds and aligned across every rendition with scene-cut keyframes disabled so segment boundaries line up and a player can switch at any segment (Live Streaming: Ingest & Transcoding).
• Ingest edge: the point of presence where a creator's encoder connection lands, close to the creator, which terminates the protocol and forwards the stream to an origin (Live Streaming: Ingest & Transcoding).
• Ingest routing: the per-stream decision, made at connect and again when capacity changes, that sends a stream from its edge to an origin with spare transcoding and backbone capacity (Twitch's Intelligest) (Live Streaming: Ingest & Transcoding).
• Ingest session: one encoder connection feeding a stream, recorded with its protocol, edge, origin, codecs and outcome; a stream survives several sessions through reconnects and takeovers (Live Streaming: Ingest & Transcoding).
• Ingredient (C2PA): a source file recorded in a C2PA manifest when an edit is made in a C2PA-aware tool, building a tree of provenance (Images at Scale).
• Input envelope: the maximum resolution, frame rate and bitrate a channel's tier accepts at ingest; a feed outside it is refused with a reason rather than accepted and transcoded badly (Live Streaming: Ingest & Transcoding).
• IPTC Digital Source Type: a controlled vocabulary written into image metadata describing where an image came from (digitalCapture, computationalCapture, trainedAlgorithmicMedia and others), with terms retired over time (Images at Scale).
• JPEG XL: an image format supported in Chrome 155, Edge 155, Firefox 158 and Safari 17 per MDN as of October 2026, which MDN still advises pairing with an AVIF, WebP or JPEG alternative (Images at Scale).
• Just-in-time encoding: encoding that waits for the first viewer instead of running at upload, as Mux offers, so videos nobody watches never pay for their encodes (Video on Demand).
• Latency tier: a named per-stream live latency setting (standard, reduced, low, real time) chosen before going live, each giving up something in quality ceiling, robustness, device reach, features or cost (Live Latency, Playback & VODs).
• Lazy variant: an image variant made through the pipeline on first request and then written to media_variants and storage, so the second request is a read even after the CDN evicts it (Images at Scale).
• Lead Moderator: a Twitch channel role, assignable from November 2025, that can add and remove moderators and change all moderation settings (Live Chat & the Interactive Layer).
• LL-HLS (Low-Latency HLS): HLS with partial segments, preload hints, blocking playlist reloads, delta updates and rendition reports, which brings live latency under about five seconds (Live Latency, Playback & VODs).
• Lockdown: a preset "under attack" bundle of chat settings (long followers-only, no first-time chatters, stricter holds, slow mode) applied in one click and logged as one decision; Twitch's version, Shield Mode, launched November 30, 2022 (Live Chat & the Interactive Layer).
• Long tail: the mass of rarely requested files that miss the CDN and are served from storage, the storage system's real customer and the reason on-demand variants requested there are encodes rather than reads (Storage, Dedup & Delivery Economics).
• Message review handler: Amazon IVS Chat's hook that calls the platform's function for every message and waits at most 200 ms, applying a preset fallback result (allow or deny) on a timeout or error (Live Chat & the Interactive Layer).
• Metadata stripping: removing EXIF and XMP metadata (GPS, capture time, serial numbers) from every served variant, keeping only an allowlist, after orientation and color profile have been applied to the pixels (Images at Scale).
• Mezzanine file: the high-quality master a platform transcodes from; platforms publish a mezzanine spec (YouTube's recommended upload settings) because the input sets the quality ceiling and how fast processing starts (The Upload Path).
• Minimum storage duration: the minimum period a colder storage class bills for (30, 90 or 180 days), so moving young or soon-to-be-purged files into it pays the minimum anyway (Storage, Dedup & Delivery Economics).
• Multipart upload: S3's protocol for large objects: parts numbered 1 to 10,000 sent in any order and completed with a list; incomplete uploads have no expiry and are billed until completed or aborted (The Upload Path).
• Multivariant playlist: the top-level HLS playlist (the RFC's master playlist) listing each variant stream with its peak BANDWIDTH, codecs and resolution, plus alternate audio, subtitle and caption renditions (Video on Demand).
• Near match: a perceptual near-duplicate pair within the platform's distance threshold, stored as a lead for repost collapsing, takedown sweeps and re-upload detection, never as an automatic verdict (Storage, Dedup & Delivery Economics).
• Non-moderator chat delay: holding every non-moderator chat message for a few seconds (Twitch offers 2, 4 or 6) so moderators and bots can remove it before viewers see it (Live Chat & the Interactive Layer).
• On-demand variant: an image variant made the first time a URL asks for it and cached afterwards, so redesigns need no backfill, at the risk of an encoding storm when the cache is cold (Images at Scale).
• Opaque ID: the viewer identifier an extension receives instead of the account (on a platform's own build, a keyed hash of extension and account, so it can't be joined across extensions); the real ID is released only after the viewer accepts an identity-sharing prompt (Live Chat & the Interactive Layer).
• Origin (live): where a live stream's video is transcoded, chosen for spare transcoding capacity and backbone headroom rather than nearness to the creator (Live Streaming: Ingest & Transcoding).
• Orphan: stored data nothing points at, such as abandoned uploads, unattached assets, variants of purged originals and anonymous uploads, which a media pipeline produces constantly (Storage, Dedup & Delivery Economics).
• P2P assist (peer-to-peer assist): viewers sharing video segments with each other to offload the CDN, at the cost of viewers' upload bandwidth and a more complex player; Twitch tested it in Korea (Live Latency, Playback & VODs).
• Partial segment: an LL-HLS subdivision of a segment (EXT-X-PART) no longer than a part target, which a player can request before the full segment exists (Live Latency, Playback & VODs).
• Past broadcast: the recording of a live stream kept as an on-demand video for a period that, on Twitch, has historically depended on the channel's program status (Live Latency, Playback & VODs).
• Per-title encoding: computing an encoding ladder for each video instead of using a fixed table, because the right bitrate depends on the content; Netflix moved to it in December 2015 (Video on Demand).
• Perceptual hash: a hash of an image's or video's content that stays close for the same picture recompressed, resized or lightly cropped, used to find near-duplicates and to match against hash lists (Storage, Dedup & Delivery Economics).
• Perceptual quality: a quality target set per image to hold a perceived level (Cloudinary's q_auto, Cloudflare's perceptual levels) instead of one fixed number, since the same setting means different things per image and encoder (Images at Scale).
• Persistent license: a DRM license stored on the device for offline playback, with an expiry and a play window (Video on Demand).
• Pipeline version: the version of encoder settings recorded on every variant, so a change in quality, format or stripping rolls out by regenerating stale variants in the background rather than invalidating everything (Images at Scale).
• Platform events: live state updates (a poll's tally, a goal's progress, a viewer count) distributed on a separate path from chat lines and coalesced to the latest value, since a client only needs the newest state (Live Chat & the Interactive Layer).
• PlayReady: Microsoft's DRM system, one of the three (with Widevine and FairPlay) that cover the market (Video on Demand).
• Poll (live): a question with 2 to 5 choices run in a channel for a fixed window, one at a time, where viewers vote once and a channel may allow extra votes paid in channel points (Live Chat & the Interactive Layer).
• Prediction: a question with 2 to 10 outcomes on which viewers stake channel points during a window; the broadcaster locks, resolves (paying those who picked the winner) or cancels it with refunds, and Twitch cancels it after 24 hours unresolved (Live Chat & the Interactive Layer).
• Presigned URL: a URL signed to grant one storage operation until it expires, which Amazon calls a bearer token: reusable until expiry (up to 7 days from the SDKs) and bound to the credentials that signed it (The Upload Path).
• Probe: the record made of every video on arrival (container, codecs, resolution, frame rate, bitrate, fast start), used to accept fixable deviations, reject the rest and schedule the transcode (The Upload Path).
• Processing tier: which server-side processing a channel gets (guaranteed, priority, as available or source only) when transcoding capacity is rationed, recorded with the stream (Live Streaming: Ingest & Transcoding).
• QoE (quality of experience): per-playback measurements players report (whether playback started, startup time, rebuffering, rungs played, upscaling), measured per rendition to find a wrong ladder or device rule (Video on Demand).
• Quarantine: the media asset state a hash-list match moves an upload into: the original is kept in restricted storage under a legal hold and a moderation case opens, in one transaction, rather than deleting evidence (The Upload Path).
• Raid: a broadcaster sending their viewers to another live channel at the end of a stream after a countdown (90 seconds on Twitch), with the target's consent setting and the sender accountable for what arrives (Live Chat & the Interactive Layer).
• Real-time stage: a small set of participants talking in real time over WebRTC through an SFU (IVS: up to 12 hosts, under 300 ms), with a composition broadcast to the audience at a higher latency tier (Live Latency, Playback & VODs).
• Reconnect window: the time a platform holds a stream open after the encoder disconnects, so a reconnect continues the same playlist and recording (Mux: 60 seconds by default for standard latency, adjustable up to 1,800) (Live Streaming: Ingest & Transcoding).
• Recording merge: writing an encoder's reconnect within a window into the same recording when resolution, frame rate and codecs match, so a broadcast yields one recording rather than a fragment per Wi-Fi drop (Live Latency, Playback & VODs).
• Referrer-limited delivery: a CDN rule denying image requests whose Referer doesn't name an allowed domain; easy to bypass, and it also hides images from image search and link previews (Storage, Dedup & Delivery Economics).
• Rejected (asset state): the terminal state for an upload that failed validation or whose session expired: its bytes are deleted and the row stays with a rejected_reason (The Upload Path).
• Rendition: one encoded version of a video at a given resolution, bitrate and codec, stored as a row in media_variants; a rung of the encoding ladder (Video on Demand).
• Rendition (C2PA): the C2PA term for a copy made by non-editorial transformations such as re-encoding or scaling; a platform that signs as a claim generator can record the original as an ingredient of its variant (Images at Scale).
• Repackaging: rewriting existing encodes into new containers and manifests instead of re-encoding; Instagram cut the cost of its basic adaptive encodings 94% this way (Video on Demand).
• Replication factor: the stored bytes per byte of data; Meta's Haystack ran an effective 3.6 (three copies on RAID-6), which f4's erasure coding cut to 2.8 or 2.1 for warm data (Storage, Dedup & Delivery Economics).
• Replication tree: Twitch's hierarchy of data centers in which a viewer's request goes to the nearest one and is forwarded up until a node has the stream, collapsing a million identical requests into one per node (Live Latency, Playback & VODs).
• Repost collapsing: folding reposts of the same image in a feed or search result into one, crediting the earliest upload, using perceptual near matches (Storage, Dedup & Delivery Economics).
• Request coalescing (edge): an edge making one upstream fetch per segment however many viewers ask at once, which live delivery needs when a raid or a final multiplies an audience in a minute (Live Latency, Playback & VODs).
• Reserve and reconcile: holding a declared size or duration against an account's quota when an upload is granted, releasing the rest at confirm and everything if the session expires, as Cloudflare Stream does with maxDurationSeconds (The Upload Path).
• Resumable session (Google-style): the YouTube and Google Cloud Storage upload protocol in which a session URI receives chunked PUTs and a 308 Resume Incomplete reports the confirmed range; the session URI itself acts as an authentication token (The Upload Path).
• Resumable upload: an upload protocol that survives a dropped connection by letting the client ask the server how much it received and continue from that offset: tus, Google-style resumable sessions and S3 multipart (The Upload Path).
• Retention purge: a rule applied by a job that removes assets after a period, in a stated order, with exemptions, notice and skips for holds, recorded asset by asset (Storage, Dedup & Delivery Economics).
• RTMP and RTMPS: the legacy TCP contribution protocol every encoder and platform speaks; RTMPS adds TLS, encrypting the stream key and the video (Live Streaming: Ingest & Transcoding).
• Save-Data: a request header (Save-Data: on) a browser sends when the user asked to reduce data use; experimental and not Baseline per MDN, and any response shaped by it must list it in Vary (Images at Scale).
• Scanning gate: the scanning state between an upload being received and anything being published, which an asset leaves only when every blocking scan (validation, exact hash, hash lists and the like) has passed (The Upload Path).
• SCTE-35: splice signals marking where an ad break goes out and comes back in, carried as markers in recorded VOD playlists and mapped by the HLS draft into EXT-X-DATERANGE attributes (Live Latency, Playback & VODs).
• Shared chat: a combined chat across collaborating live channels in which each message keeps its source channel and is moderated under that channel's rules and moderators; the room is a view, the message belongs to one community (Live Chat & the Interactive Layer).
• Shoutout: a broadcaster recommending another channel in chat, a lighter form of raid, rate-limited on Twitch to once every 2 minutes and to the same channel once every 60 minutes (Live Chat & the Interactive Layer).
• Signed playback: replacing a video's ID with a short-lived signed token, minted per viewer session after the entitlement check and verified on every manifest and segment request, optionally with country and IP rules (Video on Demand).
• Signed URL: a URL (or cookie) carrying an expiring signature the CDN verifies, used for private, members-only and paid media; CloudFront recommends signed cookies for sets of files such as HLS segments (Storage, Dedup & Delivery Economics).
• Simulcasting: broadcasting one stream to several services at once, which Twitch allows on conditions unless an exclusivity agreement applies, and which some vendors offer as a restreaming feature (Live Streaming: Ingest & Transcoding).
• Soft binding: an invisible watermark or fingerprint that lets a C2PA credential be found again after it has been stripped from the file (Images at Scale).
• Source cap: a ceiling on source quality applied at ingest where delivery is expensive, as Twitch capped source at 720p in South Korea before closing its business there on February 27, 2024 (Live Streaming: Ingest & Transcoding).
• SRT: a UDP contribution protocol that retransmits lost packets within a latency window, drops late ones, encrypts with AES and carries a stream ID; its IETF draft is an expired individual submission, not a standard (Live Streaming: Ingest & Transcoding).
• SSAI (server-side ad insertion): the platform signaling an ad break and an ad server stitching ad segments into each viewer's playlist; on IVS it requires the MPEG transport stream container (Live Latency, Playback & VODs).
• Storage cap: a limit on what a free account or channel keeps (Flickr's 1,000 photos and videos from 2019, Twitch's 100 hours of highlights and uploads from 2025), set with an explicit order, exemptions and notice (Storage, Dedup & Delivery Economics).
• Storyboard: a sprite sheet of frames at regular intervals with a WebVTT or JSON index of tile positions and times, which drives the preview on timeline hover (Video on Demand).
• Stream key: a long-lived bearer credential that lets whoever holds it broadcast on a channel: one active key per channel, stored as an HMAC for lookup, revealed only after step-up and rotatable in one click (Live Streaming: Ingest & Transcoding).
• Stream takeover: Amazon IVS's mechanism in which a new connection with a higher priority on the stream key replaces the running one without disconnecting viewers or splitting the recording (Live Streaming: Ingest & Transcoding).
• Suspicious user: a Twitch chatter status (ACTIVE_MONITORING or RESTRICTED) that flags or holds a possible ban evader's messages without banning a possibly innocent account (Live Chat & the Interactive Layer).
• Takedown sweep: queuing the perceptual near matches of a copy removed for a policy or legal reason for review, not automatic removal (Storage, Dedup & Delivery Economics).
• Timed metadata: data placed at a point in the video rather than in wall-clock time (ID3 tags in segments on IVS, EXT-X-DATERANGE in HLS) so every viewer gets it at the same moment of the stream; it should carry an identifier, not content (Live Latency, Playback & VODs).
• Tombstone (URL): the response a removed file's URL returns afterwards (HTTP 410, or a placeholder image for embeds) so caches stop refilling from the origin and embeds show a clear removal (Storage, Dedup & Delivery Economics).
• Transcode farm: the job system that encodes video: a queue of work units (chunks cut at keyframes), a pool of CPU, GPU or ASIC encoders, and a scheduler that gives first playable renditions priority over backfills (Video on Demand).
• Transcoding (live): decoding a creator's live feed and re-encoding it into several renditions at lower resolutions and bitrates, the largest compute bill of a live platform (Live Streaming: Ingest & Transcoding).
• Transformation spec: the canonical, ordered form of an image variant's parameters (width snapped to an allowlist, fit, quality), so two URLs that mean the same transform hit the same row and cache key (Images at Scale).
• Transmuxing: repackaging a creator's encoded video into the delivery format (RTMP in, HLS segments out) without re-encoding: almost free in compute, but viewers get exactly the bitrate the creator sent (Live Streaming: Ingest & Transcoding).
• tus: an open resumable upload protocol: the client creates an upload with POST, asks the server's offset with HEAD and continues with PATCH from exactly that offset, with extensions for expiry, checksums, termination and concatenation (The Upload Path).
• Unique transformation: a distinct image transform Cloudflare bills per calendar month ($0.50 per 1,000 beyond the included 5,000, as of October 2026), excluding format and quality, one reason widths are snapped to an allowlist (Images at Scale).
• Upload session: one grant to send bytes for one asset, carrying the protocol, the byte and duration caps from the purpose and plan, the server's received offset and an expiry after which a sweeper cleans up (The Upload Path).
• Variant: a derived file of a media asset (a size, crop, format and quality made from the original), the only thing a platform serves; images, posters and video renditions all live in media_variants (Images at Scale).
• Vary: Accept: the response header telling caches a response depends on the request's Accept header, required for server-side format negotiation and best normalized into a few capability buckets (Images at Scale).
• VCU (video coding unit): Google's Argos transcoding ASIC, deployed across tens of thousands of servers, giving 8x (H.264) to 20x (VP9) the throughput of software for offline two-pass encoding (Video on Demand).
• Verified bot: a Twitch bot status that raises chat limits (7,500 messages per 30 seconds, 2,000 joins per 10 seconds) but doesn't exempt the bot from AutoMod or grant the chat bot badge (Live Chat & the Interactive Layer).
• VP9: Google's video codec, which YouTube's VCU accelerates alongside H.264, absent from Apple's HLS authoring specification (Video on Demand).
• Warm tier: millisecond-retrieval storage classes with a 30-day minimum and a retrieval fee (S3 Standard-IA, R2 Infrequent Access), for originals of older media that must still load on request (Storage, Dedup & Delivery Economics).
• Warning (chat): a Twitch moderation action that blocks a user from chatting until they acknowledge it, with a reason of up to 500 characters, an early rung below timeouts and bans (Live Chat & the Interactive Layer).
• Watch-time gate: running expensive encodes (per-title ladders, better codecs) only once a video's projected watch time crosses a threshold, so videos nobody watches never pay for them (Video on Demand).
• WebP: an image format supported by every current major browser, the safe modern default for still and animated images (Images at Scale).
• WHIP (WebRTC-HTTP Ingestion Protocol): RFC 9725 (March 2025): publishing a WebRTC stream with an SDP offer in an HTTP POST and a bearer token, allowing in-browser go-live with one audio and one video track per session (Live Streaming: Ingest & Transcoding).
• Widevine: Google's DRM system, built on Common Encryption with a hardware root of trust provisioned at the factory (Video on Demand).

Creators and Money#

• Ad load: the number of ads served per session or surface, the main advertising dial, with diminishing returns: Pinterest served 49% more ads in 2025 while the average price per ad fell 22% (Monetization Strategies).
• Ad repository: the searchable record of served ads a very large online platform must publish under DSA Article 39, kept until a year after each ad last ran, with the advertiser, payer, period, targeting parameters and reach (Ads, Brand Safety & Data Licensing).
• Ad revenue share: a creator's percentage of the net revenue the platform earned from ads on the creator's content, such as YouTube's 55% on watch-page ads (Creator Monetization).
• Ad review: the check of an ad against ad policy (prohibited and restricted categories, claims, landing pages, targeting rules) before its first impression, with violations accruing to the advertiser's account (Ads, Brand Safety & Data Licensing).
• Ad transparency (DSA Article 26): the duty to let each recipient see in real time that something is an ad, on whose behalf it runs, who paid and the main targeting parameters, and the ban on ads profiled from special categories of personal data (Ads, Brand Safety & Data Licensing).
• Ads Incentive Program: Twitch's ad share, which from April 2023 paid opted-in Partners 55% of net ad revenue when the channel ran at least three ad minutes an hour and 30% below that (Creator Monetization).
• Affiliate commission: a creator's earnings from a merchant's sale attributed to a tagged product, at a seller-set rate computed on the day of purchase and reversed on return (Creator Monetization).
• Allowlist: a list of pre-approved communities an advertiser runs on, excluding every other; on a community platform the community is the unit of allowlists and exclusions (Ads, Brand Safety & Data Licensing).
• Attribution window: the period after an impression or click within which a sale or booking is credited to a promotion, such as the 30 days eBay and Etsy's Offsite Ads use (Monetization Strategies).
• Audience threshold: an eligibility gate set on audience size, such as YouTube's 1,000 subscribers plus 4,000 watch hours in 12 months or 10 million Shorts views in 90 days (Paying Creators & Sellers).
• Backup withholding: the 24% US withholding a platform applies to payouts when a payee hasn't given valid tax information (Paying Creators & Sellers).
• Bits: Twitch's virtual currency, bought by viewers in packs and spent to cheer, paying the creator one cent per Bit used (Creator Monetization).
• Brand safety: the floor of content no advertiser should fund at all, whatever its controls (Ads, Brand Safety & Data Licensing).
• Brand Safety Floor: GARM's per-category list of content no ad should fund, such as child sexual abuse material, graphic promotion of terrorism, piracy and malware (Ads, Brand Safety & Data Licensing).
• Brand suitability: the band of sensitive content above the brand safety floor that some brands accept and others don't, chosen per brand or campaign (Ads, Brand Safety & Data Licensing).
• Breakage: currency units bought and never spent, recognized as revenue once the platform can estimate them (Creator Monetization).
• Brokered sponsorship: a brand deal the platform arranges for a creator, which makes the platform a party to the money with a split, a contract and earnings like any other program (Creator Monetization).
• Catalog ad: an ad generated from an advertiser's product feed, such as Reddit's Dynamic Product Ads or Pinterest's shopping formats (Ads, Brand Safety & Data Licensing).
• Clawback: recovery of money already paid out when a refund, chargeback or claim arrives later, in a published order: next earnings, the payout account, then collections or write-off (Paying Creators & Sellers).
• Commercial communication declaration: the DSA Article 26(2) function letting a user declare that their content is or contains commercial communications, after which the platform labels it for others in real time (Ads, Brand Safety & Data Licensing).
• Contextual targeting: choosing an ad from the content being viewed (the community, the query, the thread) rather than a per-person profile, which avoids the consent dependency for the targeting itself (Ads, Brand Safety & Data Licensing).
• Contributor Fund: Shutterstock's pooled fund that pays contributors a share of dataset licensing revenue, at a 20% average royalty allocated by the volume of their included content (Ads, Brand Safety & Data Licensing).
• Contributor Program: Reddit's program, rolled out in the US from September 2023, converting Gold users receive into cash, at $0.90 or $1.00 per gold by karma tier as reported at launch (Creator Monetization).
• Conversation-page ad: an ad on a post's own page, above or among the comments, where logged-out search traffic lands, so the post's suitability label is what matters (Ads, Brand Safety & Data Licensing).
• Cost per sale: promoted-listing pricing that charges the seller a percentage of the sale only when a buyer clicks the promoted listing and buys within the attribution window; cost per click charges for traffic instead (Monetization Strategies).
• Creator Pool: the part of a country's monthly Shorts Feed ad revenue left after music licensing, allocated to monetizing creators by share of engaged views, of which the creator keeps 45% (Creator Monetization).
• Creator subscription: a recurring payment from a fan to a creator or community at a price from a fixed menu, with perks the platform delivers, where every renewal also creates earnings (Creator Monetization).
• DAC7: the EU directive (2021/514) requiring platform operators to collect, verify and report data on sellers of property rentals, personal services, goods and transport rentals, by 31 January for the prior year (Paying Creators & Sellers).
• Data license: a contract for structured access to the corpus, specifying scope, exclusions, freshness, deletion propagation, attribution, fees and term, and audit rights (Ads, Brand Safety & Data Licensing).
• Deletion propagation: the license term for how soon a deletion, removal, account closure or new opt-out must reach the licensee and what the licensee must then do with it (Ads, Brand Safety & Data Licensing).
• Demonetization: removal or limitation of a creator's ad earnings on content or a channel under advertiser-friendly rules, a decision that needs a notice and an appeal even though nothing was taken down (Ads, Brand Safety & Data Licensing).
• Direct support: money a fan pays for a specific creator (subscriptions, gifts, tips, paid messages), where the creator's share is a percentage or fixed amount of an identifiable payment (Creator Monetization).
• Earned units: currency a creator received, such as Reddit Gold, held as a liability to the creator and convertible to cash at a published rate with a minimum (Creator Monetization).
• Effective ad label: the most restrictive of the content, community and account suitability labels in force when an ad was served (Ads, Brand Safety & Data Licensing).
• Eligibility gate: a rule deciding who can earn at all before any money moves: an audience threshold, account integrity, or identity and tax before supply (Paying Creators & Sellers).
• Endorsement Guides: the FTC's rules (16 CFR Part 255, revised 2023) treating free products and commissions as connections creators must disclose clearly, repeated during livestreams (Creator Monetization).
• Excluded seller: a seller DAC7 doesn't require reporting, such as government entities, listed companies, property sellers with over 2,000 rentals of one address and goods sellers with fewer than 30 sales under €2,000 (Paying Creators & Sellers).
• Feed ad: a sponsored item in a feed or dashboard styled like the posts around it, whose suitability is judged per position because neighbors change on every refresh (Ads, Brand Safety & Data Licensing).
• Form 1042-S: the US information return for US-source income paid to foreign persons, reporting the gross and the tax withheld (Paying Creators & Sellers).
• Form 1099-K: the US information return for payments a platform settles as a third-party settlement organization for goods or services, filed only when a payee's gross payments exceed $20,000 and 200 transactions (the threshold reinstated in 2025) (Paying Creators & Sellers).
• Form 1099-MISC: the US information return for amounts a platform pays a US creator as its own payer, such as an ad share or pool allocation (or Form 1099-NEC) (Paying Creators & Sellers).
• Framework version: the named suitability vocabulary a label was stored against, kept because industry categories change and labels must be remapped (Ads, Brand Safety & Data Licensing).
• Freemium: free to read, free to post, paid to stand out or to go deeper; the pattern for platforms whose content is abundant and contributed (Monetization Strategies).
• Freshness: the license term for how soon new content must reach the licensee, which becomes the platform's API performance requirement (Ads, Brand Safety & Data Licensing).
• GARM (Global Alliance for Responsible Media): the World Federation of Advertisers' initiative (2019 to August 2024) whose brand safety floor and suitability framework remain the vocabulary vendors and platforms map to (Ads, Brand Safety & Data Licensing).
• Gift subscription: a subscription paid for by one account and held by another, a named recipient or a random chat member, usually bought in bundles (Creator Monetization).
• Hard paywall: access sold to content that is scarce and owned, such as a creator's back catalog or course, as opposed to freemium perks on abundant contributed content (Monetization Strategies).
• Hype Chat: Twitch's paid chat message, launched June 2023 with streamer-set prices and a 70/30 split, deprecated five months later (Creator Monetization).
• Incentive pay: earnings with no fan payment behind them, such as an hourly rate for streaming or a transition top-up, funded from the platform's budget with explicit conditions and an end date (Creator Monetization).
• Instant payout: a priced, prefunded advance of unsettled money to a debit card or real-time rail, such as eBay's $2.00 Express payout (Paying Creators & Sellers).
• Invalid traffic: fake views and engagement whose earnings a platform may withhold, change or offset, including from future balances after payment (Creator Monetization).
• Inventory tier: an advertiser-facing choice of how much sensitive content to run beside, such as YouTube's Maximum, Moderate and Limited or Reddit's expanded, standard and limited (Ads, Brand Safety & Data Licensing).
• Limited ads: the middle advertiser-friendly outcome for a video, which "will receive limited ad earnings," between full earning and none (Ads, Brand Safety & Data Licensing).
• Link-out: a link from a mobile app to the developer's own web checkout, allowed in the US for Apple since May 2025 and for Google Play after Epic v. Google, with Apple's commission on it still litigated as of October 2026 (Monetization Strategies).
• Local subscription pricing: subscription prices set by the viewer's location and payment method, as Twitch introduced from 2021, with creators' baseline revenue protected through the transition (Creator Monetization).
• Model collision: a conflict between two revenue models run together, such as ad load against premium or take rate against leakage, which needs a written answer before the second model launches (Monetization Strategies).
• Music cost share: the part of music licensing cost deducted from creator earnings by rule, such as the Shorts pool's music share or Twitch's 50/50 DJ split (Creator Monetization).
• Negative balance: what a payee owes the platform after a refund, chargeback or claim arrives after payout (Paying Creators & Sellers).
• Opt-out: a dated row with a scope and a subject (an account or a community) excluding content from licensed exports and pushed to licensees through the deletion feed (Ads, Brand Safety & Data Licensing).
• Paid chat: paying to have a message pinned, highlighted or held in live chat, a message that is still moderated and stored with the payment (Creator Monetization).
• Paid promotion disclosure: a creator's declaration of a sponsorship or commercial relationship, which adds a label to the content and is kept with the revision it applied to (Creator Monetization).
• Paid units: currency bought with money, a liability to the buyer until spent and revenue only as it is used (Creator Monetization).
• Payee: the legal person or entity behind a creator's channel, host's listing or shop, a sub-merchant whose identity, bank account and tax status must be established before money leaves (Paying Creators & Sellers).
• Payment float: money held between the buyer's payment and the seller's payout, which earns interest for the platform; a consequence of payout triggers and holds (Monetization Strategies).
• Payout minimum: the balance a payee must reach before a payout is sent, with smaller balances rolling over (Paying Creators & Sellers).
• Payout readiness: a merchant's payout status derived from its requirements (identity verified, bank account active, tax profile current, no DAC7 data outstanding) and recomputed on every change (Paying Creators & Sellers).
• Payout release rule: a rule naming the recorded event that releases money (month close, check-in, handoff completed, buyer window closed) and an offset, stored as data (Paying Creators & Sellers).
• Payout trigger: the recorded event after which the platform considers money earned by the payee and safe to send, such as month close, check-in, trip start plus 48 hours or delivery (Paying Creators & Sellers).
• Plus Points: Twitch's measure for subscription split levels, one point per Tier 1 recurring paid sub, two per Tier 2 and six per Tier 3; 100 points for three months earn 60/40 and 300 earn 70/30 (Creator Monetization).
• Pooled program: a program that divides revenue the platform earned from someone else (advertisers, all subscribers) by each creator's measured contribution after the period closes (Creator Monetization).
• Post-bid measurement: suitability scoring of served impressions after the fact by a verification vendor, which needs a log of the content revision beside each ad (Ads, Brand Safety & Data Licensing).
• Pre-bid control: a suitability control applied before purchase, which needs the platform's labels available at auction time (Ads, Brand Safety & Data Licensing).
• Program enrollment: a payee's admission to a revenue program, recorded as a decision with criteria, decider, reason, appeal deadline and what happens to earnings on exit (Paying Creators & Sellers).
• Progressive onboarding: collecting the minimum to accrue earnings, then identity, bank and tax before the first payout, and business verification above a volume line (Paying Creators & Sellers).
• Promoted listing: paid placement of a seller's listing in the marketplace's own search and recommendations, charged per sale or per click; effectively a second, opt-in take rate (Monetization Strategies).
• Promotional units: currency granted free, with no money behind it, whose spending either creates a platform-funded creator payout or none, as the terms say (Creator Monetization).
• Protection margin: the margin a rental marketplace earns by pricing host and guest protection above expected claims, real only after claims, appraisal, fraud and subrogation (Monetization Strategies).
• Regional pricing: prices set by country to follow ability to pay, priced by the payment method's country as well as the account's (Monetization Strategies).
• Revenue program: one way money is split (a creator subscription share, an ad share, a stay service fee, a data license), with a payer role, a payee role and versioned terms (Monetization Strategies).
• Revenue program terms: dated rows saying what each party keeps, on what base, for which tier, country and channel, resolved when money is earned and stored on every earning (Monetization Strategies).
• Risk hold: a dated delay on a payout the trigger already released, for review, account, fraud, claim or new-payee reasons, often the outcome of a trust and safety case (Paying Creators & Sellers).
• Self-certification (ad suitability): a creator's answers at upload about what their content contains, a signal that can speed eligibility but not a verdict (Ads, Brand Safety & Data Licensing).
• Self-gifting: a creator or friends buying subscriptions and tips for the creator's own channel to reach a points threshold or eligibility bar, also called wash support (Creator Monetization).
• Seller subscription: a monthly plan sellers pay for tools, credits and limits, such as Etsy Plus at $10 a month (Monetization Strategies).
• Share base: the term saying what a payee's percentage is computed on: the gross price, the price net of tax, or the price net of the store's fee (Monetization Strategies).
• Shoppable ads: ad formats built from merchants' product catalogs, where the merchant pays as an advertiser and the platform matches products and measures conversions (Monetization Strategies).
• Spread: the difference between what a viewer pays for virtual currency and what the creator receives when it is spent, which is the platform's margin (Creator Monetization).
• Store-fee bearer: the term saying who absorbs an app store's fee on a purchase: the platform, the payee, both, or the payer through a higher price (Monetization Strategies).
• Store-fee offset: an in-app price raised above the web price to offset the app store's fee, recorded as its own price row (Monetization Strategies).
• Subscription period: one renewal of a subscription recorded with its price, channel, terms and points, so later changes never rewrite what was earned (Creator Monetization).
• Suitability tiers: GARM's high, medium and low risk levels per content category, from glamorized depiction (high) to educational treatment (low) (Ads, Brand Safety & Data Licensing).
• Take rate: a marketplace's revenue over its gross volume, such as Airbnb's 13.4% of gross booking value in 2025 (Monetization Strategies).
• Tax interview: the questions a payee answers before the first payout, ending in a W-9, W-8BEN or W-8BEN-E (Paying Creators & Sellers).
• Tax profile: a payee's effective-dated tax status from the tax interview, which each withholding records it was computed under (Paying Creators & Sellers).
• TDM reservation: a rightholder's machine-readable reservation against text and data mining under Article 4 of the EU Copyright in the Digital Single Market Directive, which takes the work out of the mining exception (Ads, Brand Safety & Data Licensing).
• Third-party settlement organization: a platform that settles payments for goods or services between other parties, and so files Form 1099-K for its payees (Paying Creators & Sellers).
• Tip: a one-time payment to a creator, usually attached to a chat message, comment or award, sold as money or as virtual currency (Creator Monetization).
• UK platform reporting: the UK's own reporting rules for digital platforms, with operators registering with HMRC and a goods small-seller exclusion like DAC7's (Paying Creators & Sellers).
• Unit ledger: a double-entry ledger balanced in currency units, whose paid entries carry the payment and money value behind them so refunds claw back exactly those units (Creator Monetization).
• US-source withholding: withholding of 0% to 30% by tax treaty on a non-US creator's earnings from US viewers, which is why every earning carries the viewer's country (Paying Creators & Sellers).
• W-8BEN: the certificate of foreign status a foreign individual gives a US payer, with or without a treaty claim; W-8BEN-E is the version for foreign entities (Paying Creators & Sellers).
• W-9: the US tax form giving a US person's name and taxpayer identification number to the payer (Paying Creators & Sellers).

Peer-to-Peer Marketplaces#

• Actual cash value (ACV): the value basis a total loss is paid on, up to the plan's cap ($300,000 at Outdoorsy, $200,000 at Turo) (Handoffs, Damage & Claims).
• Add-on: an optional extra a guest can add to a booking, such as a bike rack, generator, linens or delivery, defined on the listing and priced in the quote (Listings, Search & Ranking).
• Advance notice: the listing rule for how soon before the start a booking may be made, such as same-day, one day or three days (Availability & Booking).
• AirCover: Airbnb's protection brand: for guests, rebooking or a refund when the host cancels, can't be reached or the listing is significantly different; for hosts, damage protection as a guaranty plus liability insurance (Cancellations, Guarantees & Disputes).
• All-in price: the displayed total including every mandatory fee the business knows and can calculate, shown more prominently than other pricing information (Pricing, Fees & All-In Prices).
• Appraisal: the valuation of vehicle damage: a repair facility's estimate reviewed by an appraiser (Handoffs, Damage & Claims).
• Asset eligibility: the per-listing check of the thing offered (value, age, title, registration, inspection, recalls, anti-theft), with rules set mostly by the insurer or protection plan (Supply Onboarding & Verification).
• Asset record: a vehicle or property kept as its own record (VIN, year, class, value, title, lienholder) independent of any listing, so its history survives a relist or a change of host (Supply Onboarding & Verification).
• Authenticity Guarantee: eBay's program routing luxury goods and collectibles through an authenticator before they reach the buyer, turning a dispute into a pre-delivery check (Marketplace Fraud & Risk).
• Availability-aware search: rental search that returns only listings free for the requested dates by joining text, geography and the calendar, then re-checks the top results against the system of record (Listings, Search & Ranking).
• Average-consumer test: DSA Article 6(3): the hosting safe harbor doesn't cover consumer-protection liability when an average consumer would believe the item comes from the platform itself (Supply Onboarding & Verification).
• Background check: a consumer report on a user, run by some marketplaces for certain users, carrying notice, consent and adverse action duties in the US when used to deny someone (Marketplace Fraud & Risk).
• Blocked property: property of a sanctioned person in a US person's possession, frozen rather than seized and reported to OFAC within 10 business days (Marketplace Fraud & Risk).
• Bookability: the listing's own status (active, snoozed, sold out, retired), kept separate from the content item's visibility (Listings, Search & Ranking).
• Booking change: a modification (extension, shortening, delivery, added driver) recorded as a request priced against the frozen quote, agreed by the right party and applied in one transaction (Availability & Booking).
• Booking participant: each driver or guest on a booking, verified individually before the trip, the list that coverage checks read (Availability & Booking).
• Booking restriction: a published rule narrowing what a risky guest may book, such as Airbnb's limit on local entire-home bookings by US and Canadian guests under 25 (Marketplace Fraud & Risk).
• Booking states: the booking vocabulary from requested and pending payment through confirmed, in progress and completed, or ending cancelled, declined or expired (Availability & Booking).
• Booking window: the listing rule for how far ahead the calendar opens to bookings (Availability & Booking).
• Buyer protection fee: a mandatory fee charged to the buyer that funds refunds for lost, damaged or not-as-described items, as Vinted charges for purchases from non-professional sellers (Deposits, Protection Plans & Insurance).
• California Honest Pricing Law: SB 478, effective July 1, 2024, making it unlawful for most businesses to advertise a price that leaves out mandatory fees, other than government charges and reasonable shipping (Pricing, Fees & All-In Prices).
• Cancellation policy catalog: the platform-written set of named, time-based refund schedules (Flexible, Moderate, Firm, Strict and so on) a host chooses from, versioned and frozen on each booking (Cancellations, Guarantees & Disputes).
• Car sharing period: the statutory period that begins and ends at defined events (the booking's start, the handoff, the return), during which coverage must be primary and records kept (Deposits, Protection Plans & Insurance).
• Case lifecycle: the shared shape of guarantee and dispute cases: contact the other side, wait, escalate, decide, remedy, appeal (Cancellations, Guarantees & Disputes).
• Cash-out: a host booking their own listing with stolen cards, collecting the payout and disappearing before the chargebacks arrive (Marketplace Fraud & Risk).
• Claim: a request by one side, usually the host, for money for a loss tied to a booking, such as damage, theft, total loss, cleaning, usage or liability (Handoffs, Damage & Claims).
• Claim window: the filing deadline for a claim, measured from the trip's end and set by the protection in force at booking, such as 48 hours at Outdoorsy or 14 days at Airbnb (Handoffs, Damage & Claims).
• Closed to arrival or departure: a listing rule allowing pickups or returns only on certain days (Availability & Booking).
• Cold-start boost: a temporary ranking lift for new listings, which have no reviews or bookings yet, expiring by age or first bookings and measured like any feature (Listings, Search & Ranking).
• Collusion: a host and guest agreeing on a fake claim or a fake non-return to collect from the platform's protection or an insurer (Marketplace Fraud & Risk).
• Compliance by design (DSA Article 31): the duty to build the listing interface so traders can give the pre-contractual and product-safety information EU law requires, checked with best efforts before the listing goes live (Listings, Search & Ranking).
• Constructive cancellation: a listing grossly different from its description (a double booking, a substituted property) treated as a host cancellation (Cancellations, Guarantees & Disputes).
• Contact detection: detection of phone numbers, emails, links and handles in messages, including obfuscated ones and text inside images, so they can be blocked, masked or flagged (Messaging, Reviews & Reputation).
• Contractual reimbursement: a damage protection product written as a contract to reimburse rather than an insurance policy, such as Turo's host plans, which needs no producer license unless a regulator treats it as insurance (Deposits, Protection Plans & Insurance).
• Counterparty response: the other side's turn on a claim before review, with a deadline and three exits: accept, dispute or no response (Handoffs, Damage & Claims).
• Custom quote: a priced offer a host sends a guest for specific dates, accepted instantly, created as a booking in the requested state with its own expiry (Availability & Booking).
• DAC7 onboarding fields: the seller data DAC7 requires (name, address, date of birth, TIN, VAT number, financial account, and for rentals the property's address), collected at onboarding because vehicle and property rentals have no small-seller exclusion (Supply Onboarding & Verification).
• Decline reason: a fixed-list reason a host gives for declining a request, each with a consequence such as a calendar task or a counted metric (Availability & Booking).
• Delivery radius: the distance a host will deliver an asset, which makes a listing a search candidate in places it isn't parked (Listings, Search & Ranking).
• Demand rerouting: turning a guest's unfulfilled or declined request into invitations to similar available hosts, as Outdoorsy's Smart Match does (Listings, Search & Ranking).
• Deposit hold: a card authorization placed shortly before a rental, released if nothing goes wrong and partly captured if something does; in practice a sequence of authorizations (Deposits, Protection Plans & Insurance).
• Deposit waiver: a flat fee charged at booking in place of a deposit hold, buying the deductible down to a covered amount, with its own exclusions and version (Deposits, Protection Plans & Insurance).
• Digital key exchange: Outdoorsy's in-app handoff: rental contract signed, timestamped and geotagged photos, pre-existing damage tagged and readings recorded, repeated at return (Handoffs, Damage & Claims).
• Display class: each quote line's classification as mandatory, optional, government or contingent, from which the displayed all-in total is computed (Pricing, Fees & All-In Prices).
• Disruption activation: a record that switches on an extenuating-circumstances policy for an area and time range, letting qualifying bookings cancel without host penalties (Cancellations, Guarantees & Disputes).
• Drip pricing: showing a base price and adding mandatory fees at checkout, the practice all-in price rules prohibit (Pricing, Fees & All-In Prices).
• Driver eligibility: the protection plan's conditions on who may drive (age, license class, verification), checked before pickup, with unverified drivers having no damage coverage (Deposits, Protection Plans & Insurance).
• Duplicate listing detection: matching photo hashes, normalized addresses, VINs and text to find the same property or vehicle under two hosts (Marketplace Fraud & Risk).
• Dynamic pricing recommendation: an opt-in tool that sets or suggests prices within host-set bounds, logged with its model version and outcome and never changing confirmed bookings (Pricing, Fees & All-In Prices).
• Evidence photo: a handoff photo captured in the platform's app with device time, location, server receipt time and a capture hash, never stripped of metadata or edited (Handoffs, Damage & Claims).
• Excess coverage: coverage that pays only after the guest's own insurance, becoming primary only when there is none; cheaper for the platform and slower for everyone (Deposits, Protection Plans & Insurance).
• Extenuating circumstances: events beyond anyone's control (declared emergencies, evacuations, natural disasters) under which the platform overrides cancellation policies, as in Airbnb's Major Disruptive Events Policy (Cancellations, Guarantees & Disputes).
• Fake listing: a listing for a place or vehicle that isn't for rent or doesn't exist, typically priced too low with an absent owner (Marketplace Fraud & Risk).
• Fee program: the versioned terms that answer who pays a marketplace fee, on what base, at what rate (tiers, floors, caps) and under which program, resolved at booking (Pricing, Fees & All-In Prices).
• First-dollar backstop: the car sharing law rule that the program's insurance pays from the first dollar when the owner's or driver's coverage has lapsed or is inadequate (Deposits, Protection Plans & Insurance).
• Frozen quote: the booking's price lines (who pays, who earns, display class, refundability) computed once and never recomputed, with post-trip charges added as separate lines (Pricing, Fees & All-In Prices).
• FTC Consumer Reviews and Testimonials Rule: 16 CFR Part 465, effective October 21, 2024, banning fake reviews, sentiment-conditioned incentives, undisclosed insider reviews, company-controlled review sites and review suppression (Messaging, Reviews & Reputation).
• FTC Rule on Unfair or Deceptive Fees: 16 CFR Part 464, effective May 12, 2025, requiring all-in prices for live-event tickets and short-term lodging, including vacation rentals (Pricing, Fees & All-In Prices).
• Grace period: a short window after booking in which a guest can cancel for a full refund, such as 24 hours when the trip is at least 7 days away (Cancellations, Guarantees & Disputes).
• Guarantee program: the platform's promise of a remedy when the other side fails, with a scope, a precondition such as paying on the platform, a waiting period, a deadline and a decision (Cancellations, Guarantees & Disputes).
• Guaranty: a damage protection promise that is neither a waiver nor insurance, such as Airbnb's host damage protection up to $3 million (Deposits, Protection Plans & Insurance).
• Guest protection plan: a per-trip product setting the guest's maximum responsibility for damage, such as Turo's Premier, Standard and Minimum plans (Deposits, Protection Plans & Insurance).
• Handoff: the moment control of the asset changes, which proves its condition, records the readings that price the trip and marks the start or end of the coverage period (Handoffs, Damage & Claims).
• High-volume seller: under INFORM, a seller with 200 or more sales of new or unused consumer products and $5,000 or more in revenue in any 12 months of the previous 24, computed from on-platform sales (Supply Onboarding & Verification).
• Hijacked listing: a real listing's photos and description copied by a scammer who swaps in their own contact details (Marketplace Fraud & Risk).
• Host block: days a host closes on the calendar for personal use, maintenance or any reason, stored as occupancy (Availability & Booking).
• Host cancellation fee: the penalty a host pays for cancelling a confirmed booking, at Airbnb 10%, 25% or 50% of the reservation by timing, minimum $50, withheld from later payouts (Cancellations, Guarantees & Disputes).
• Host plan tier: a host protection level that trades take rate for risk, such as Turo's plans keeping 70%, 80% or 90% of the trip price against damage responsibility of $250, $1,500 or $2,750 (Deposits, Protection Plans & Insurance).
• Host-only fee: a single platform fee charged to the host so the guest sees one price that already includes the platform's cut, such as Airbnb's 15.5% for most hosts (Pricing, Fees & All-In Prices).
• Identity graph: a graph of accounts linked by shared devices, payment instruments, documents, addresses and payout accounts, used to resolve identity and find linked accounts (Marketplace Fraud & Risk).
• Imported block: days closed by an external calendar feed, with no guest, price or contract on this marketplace (Availability & Booking).
• INFORM Consumers Act: the US law (15 U.S.C. 45f), effective June 27, 2023, requiring marketplaces to collect, verify and in some cases disclose high-volume sellers' data, with a reporting channel on their listings (Supply Onboarding & Verification).
• Instant book: booking confirmed at checkout under rules the host set, without the host approving each guest (Availability & Booking).
• Instant book eligibility: the host-chosen conditions on guest and trip (trip type, verification, review history, driver age, length) under which a booking confirms without approval, frozen on the booking (Availability & Booking).
• Insurance producer: anyone who sells, solicits or negotiates insurance, licensed by each US state, which is why a marketplace offering coverage at checkout holds licenses or sells through a licensed firm (Deposits, Protection Plans & Insurance).
• Leakage: a guest and host who meet on the platform and transact without it, costing the take rate and the guest's protections (Messaging, Reviews & Reputation).
• Lienholder: a lender with a lien on a financed vehicle, who may forbid sharing, must be noticed before listing under some car sharing laws, and is paid first on a total loss (Supply Onboarding & Verification).
• Limited payment collection agent: the role a marketplace claims when it collects payment on the host's behalf without being a party to the rental contract, as Outdoorsy's terms describe (Supply Onboarding & Verification).
• Linked accounts: accounts tied to one operator by shared evidence, read to make penalties stick and to find self-dealing, rings and payout concentration (Marketplace Fraud & Risk).
• Listing quality score: the platform's versioned, explained estimate of how likely a listing is to deliver a good trip, built from content completeness, outcome history and host behavior (Listings, Search & Ranking).
• Loss of use: the bookings a host loses while a vehicle is repaired, excluded by Outdoorsy's and Turo's host protection and included in Airbnb's for stays (Handoffs, Damage & Claims).
• Major Disruptive Events Policy: Airbnb's override for declared emergencies, travel restrictions, military action, prolonged outages and unforeseeable disasters, letting guests cancel for a refund and hosts without fees (Cancellations, Guarantees & Disputes).
• Marketplace facilitator: a marketplace that state law makes calculate, collect and remit sales or rental tax on third-party sales (Pricing, Fees & All-In Prices).
• Mass arbitration: a law firm filing thousands of individual arbitration demands at once, each carrying fees the company pays, now handled by special provider rules (Cancellations, Guarantees & Disputes).
• Merchant of record: the party whose name is on the card statement and who refunds and absorbs chargebacks, which decides who earns each fee line and who owes the tax (Pricing, Fees & All-In Prices).
• Money Back Guarantee: eBay's guarantee for items not received or not as described, with a seller response period, eBay stepping in after it, and no chargeback allowed for the same issue (Cancellations, Guarantees & Disputes).
• Not as described: a goods claim that the item received doesn't match the listing, judged against the listing revision the buyer purchased (Handoffs, Damage & Claims).
• Not-a-party posture: marketplace terms that put the rental or sale contract between the two users, a position tested by the DSA, airports, car sharing laws and tax authorities (Supply Onboarding & Verification).
• Occupancy: every claim on a listing's time (bookings, holds, host blocks, imported blocks, buffers, eligibility closures) kept in one table under one exclusion constraint (Availability & Booking).
• Onboarding gate: one of five separate checks a supplier passes (identity, payout, tax, asset, content), each with its own state, validity and the act it blocks (Supply Onboarding & Verification).
• One-remedy rule: a guarantee's ban on also taking a chargeback for the same issue, enforced by linking the case, the refund and any dispute on the payment (Cancellations, Guarantees & Disputes).
• Payment diversion: persuading a guest to pay outside the platform, often through a link to a fake copy of its checkout, where every platform protection ends (Marketplace Fraud & Risk).
• Peer-to-peer car sharing law: a state statute, such as Washington's SHB 1389, regulating the car sharing program directly: primary insurance during the sharing period, disclosures, lienholder notice, records and recalls (Deposits, Protection Plans & Insurance).
• Platform-initiated cancellation: a cancellation by the marketplace itself, such as during a wind-down, refunded in full with no penalty on either side (Cancellations, Guarantees & Disputes).
• Post-trip charge: an amount discovered after return (damage beyond the deposit, cleaning, fuel, tolls, overages) charged to the card on file with consent at booking, evidence and notice (Deposits, Protection Plans & Insurance).
• Preparation time: the buffer before and after each booking for cleaning and checks, stored inside the occupied range so the database can enforce it (Availability & Booking).
• Primary coverage: coverage that pays first, before the guest's own insurance; car sharing laws require that whichever coverage applies during the sharing period is primary (Deposits, Protection Plans & Insurance).
• Private individual: a supplier not acting for trade, business or profession, outside most consumer law, so buyers rely on the platform's guarantee (Supply Onboarding & Verification).
• Process arbitrator: under mass arbitration rules, the arbitrator who handles non-merits issues before individual merits arbitrators decide each claim (Cancellations, Guarantees & Disputes).
• Product compliance fields: typed per-category listing fields for EU product-safety information, checked before publishing and sampled against illegal-product databases (Listings, Search & Ranking).
• Promotion funder: the party whose earnings a discount reduces (the platform or the host), recorded on each promotion line with its base and stacking rule (Pricing, Fees & All-In Prices).
• Protection stack: the layers that can pay a rental loss (deposit or waiver, guest plan, host plan or guaranty, liability insurance, physical damage insurance, each party's own insurance), each with a legal form (Deposits, Protection Plans & Insurance).
• Readings: odometer, generator hours, fuel or charge level captured with a photo at both handoffs, used to price usage and as claim evidence (Handoffs, Damage & Claims).
• Rebooking: the guest's replacement booking on a comparable listing after a host cancellation, linked to the cancelled one with the price difference funded by the platform (Cancellations, Guarantees & Disputes).
• Recall feed: a periodic check of every listed vehicle against a safety recall source, which some car sharing laws require before a vehicle is made available (Supply Onboarding & Verification).
• Recommender transparency: DSA Article 27's duty to set out a ranking's main parameters and users' options in plain language in the terms (Listings, Search & Ranking).
• Reimbursement-only settlement: paying a claim only against documented repairs rather than an estimate, used where fraud is suspected (Handoffs, Damage & Claims).
• Remitter: the party a tax rule makes responsible for paying a tax line to the authority, the platform as facilitator or the host (Pricing, Fees & All-In Prices).
• Request to book: a booking the host accepts or declines within a window, with payment authorized at request and the request expiring if unanswered (Availability & Booking).
• Reservation risk score: an estimate of the chance a booking ends in a loss (fraud, misuse, damage, non-return), computed at several checkpoints from trip and payment features (Marketplace Fraud & Risk).
• Review coercion: tying a deposit, refund or discount to a review's content, a policy violation detectable in messages and potentially a violation of the FTC reviews rule (Messaging, Reviews & Reputation).
• Review suppression: removing or deterring honest reviews through false accusations, intimidation or unfounded legal threats, prohibited by the FTC reviews rule (Messaging, Reviews & Reputation).
• Roadside assistance: towing, tire, battery, lockout and fuel help bought from a provider and attached to the trip, recorded with its program and tier (Deposits, Protection Plans & Insurance).
• SDN screening: screening hosts, sellers, guests and payees against OFAC's Specially Designated Nationals list at onboarding, on changes, on list updates and before payouts, with potential matches reviewed (Marketplace Fraud & Risk).
• Self-certification (trader): the statement DSA Article 30(1)(e) requires from a trader, committing to offer only products or services that comply with EU law, collected before they can sell to EU consumers and shown to buyers; distinct from a creator's ad-suitability self-certification (Supply Onboarding & Verification).
• Self-dealing: a host booking their own listing from a second account to inflate reviews or cash out stolen cards, visible as guest and host linked by device or payment (Marketplace Fraud & Risk).
• Seller performance standards: eBay's seller metrics computed from transaction defects rather than feedback, which negative or neutral feedback doesn't affect (Messaging, Reviews & Reputation).
• Settlement order: the sequence a settled claim is paid from: the deposit, the guest's card on file, the guest's plan, the host's plan or guaranty, an insurer, then the guest's own insurer (Handoffs, Damage & Claims).
• Simultaneous reveal: holding each side's review sealed until both have submitted or the window closes, so neither can retaliate against what the other wrote (Messaging, Reviews & Reputation).
• Split payment: a booking paid in installments, such as a 50% deposit with the balance charged automatically 30 days before departure (Pricing, Fees & All-In Prices).
• Status program: a badge and perks for hosts or sellers who meet published thresholds on responsiveness, rating, reliability and volume, evaluated on a schedule, such as Superhost (Messaging, Reviews & Reputation).
• Subrogation: the payer of a host's loss stepping into the host's claim against the guest to recover from the guest, their insurer or collections (Handoffs, Damage & Claims).
• Supplement: an addition to a repair estimate when the repair uncovers more damage than first shown, reviewed by an appraiser (Handoffs, Damage & Claims).
• Supply culling: retiring listings that underperform or generate claims, with a reason, rather than deleting them (Listings, Search & Ranking).
• Supply sanction: an automatic penalty on a host for cancellations or unresponsiveness, such as Outdoorsy pausing instant book for 45 days, modeled as a penalty rather than a setting change (Availability & Booking).
• Total loss: a loss where repair costs approach the vehicle's value, paid at actual cash value up to the plan's cap with the lienholder paid first (Handoffs, Damage & Claims).
• Trader: under the DSA, someone acting for purposes of their trade, business, craft or profession, which brings collection, verification and disclosure duties (Supply Onboarding & Verification).
• Trader traceability: DSA Article 30's duty to collect a trader's identity, contact, payment and registration data and a compliance self-certification before they offer to EU consumers, verify it with best efforts and keep it six months after the relationship ends (Supply Onboarding & Verification).
• Transaction thread: a marketplace conversation tied to a listing and, once booked, a booking, with immutable messages kept as the record of what was promised (Messaging, Reviews & Reputation).
• Usage charge: a price known only after the trip (miles, generator hours, fuel, late return), computed from rates frozen at booking and readings taken at handoff, with a deadline (Pricing, Fees & All-In Prices).
• Visibility: the content item's state for a listing (draft, pending review, published, hidden, removed, deleted), kept separate from bookability (Listings, Search & Ranking).

Trust and Safety#

• 15 U.S.C. 6851: the federal civil action, since 1 October 2022, against a person who discloses an intimate visual depiction without consent, with liquidated damages of $150,000 and pseudonymous suits (Intimate Image Abuse).
• 18 U.S.C. § 2258A: the US statute requiring providers to report apparent child sexual exploitation to NCMEC's CyberTipline as soon as reasonably possible after obtaining actual knowledge, with no duty to scan (Child Safety: Detection & Reporting).
• 512(c) safe harbor: the DMCA safe harbor for content stored at users' direction, conditional on no knowledge (or expeditious removal), no direct financial benefit where the platform can control the infringement, removal on valid notice and a registered designated agent (The Law of Hosting User Content).
• 512(f): the DMCA provision making anyone who knowingly materially misrepresents that material is infringing liable for the resulting damages, costs and fees (Copyright at Scale).
• 512(h) subpoena: a subpoena a copyright owner gets from a federal court clerk, without a judge's review, ordering a platform to identify an alleged infringer (Copyright at Scale).
• Abuse contact: the published address for abuse reports: the abuse@ role mailbox from RFC 2142, a login-free web form, and the abuse entries in registration and IP registry records (Abuse Desks for Hosts & Registrars).
• Abuse desk: the team and channels through which a host, CDN or registrar receives reports about its customers from people who usually aren't its customers (Abuse Desks for Hosts & Registrars).
• Abuse mitigation: one recorded action against one resource, with its layer, target, collateral, reversibility, evidence, customer notice and the deadline given (Abuse Desks for Hosts & Registrars).
• Access software provider: in Section 230, software that can filter, screen, allow, disallow, pick, choose, analyze, digest, organize or reorganize content (The Law of Hosting User Content).
• Account Creation (OAT-019): OWASP's automated-threat name for bulk account creation, and sometimes profile population, through a site's sign-up process (Spam, Bots & Inauthentic Behavior).
• Account farm: an operation that registers accounts in bulk, ages them and builds standing so they can later be used or sold for spam, scams or manipulation (Spam, Bots & Inauthentic Behavior).
• Account link: recorded evidence that two accounts share an operator (device, payment instrument, phone, network, behavior), with strength and status; evidence, never a merge (Spam, Bots & Inauthentic Behavior).
• Actionable evidence: evidence sufficient for a registrar to make a reasonable determination that a name is used for DNS Abuse, which obliges it to promptly take appropriate mitigation (Abuse Desks for Hosts & Registrars).
• Active role: the DSA limit (Recital 18) under which a provider that goes beyond merely technical and automatic processing, gaining knowledge of or control over information, loses the hosting exemption for it (The Law of Hosting User Content).
• Actual knowledge: the trigger most hosting regimes share: once a platform knows of specific illegal or infringing material it must act, and the earliest provable receipt of a notice is when that knowledge starts (The Law of Hosting User Content).
• Age restriction (content): a gate that hides content from users under 18 and signed-out viewers, used instead of removal when context is enough (Terrorism, Violent Extremism & Graphic Content).
• Article 17: the EU DSM Directive rule that makes content-sharing services directly liable for unauthorised uploads unless they show best efforts to license, to prevent availability of identified works, and to act on notices and prevent re-uploads (Copyright at Scale).
• Audible Magic: an automated content identification vendor whose audio recognition Twitch uses on VODs (Copyright at Scale).
• Authentic intimate visual depiction: a real intimate image of an identifiable adult; publishing one is a federal offense under the TAKE IT DOWN Act when the privacy, public-concern and harm conditions are met (Intimate Image Abuse).
• Back-scan: running each new batch of victim-submitted hashes against media already stored, because the image may have been posted before it was hashed (Intimate Image Abuse).
• Bellwether: a first trial chosen from a large set of coordinated cases; K.G.M. was the first social media addiction bellwether to reach a jury verdict, in March 2026 (The Law of Hosting User Content).
• Best efforts: the Article 17 standard a content-sharing service must show to escape liability: efforts to obtain authorization, to keep identified works unavailable under high industry standards, and to prevent re-uploads (Copyright at Scale).
• Blocklist (abuse reputation): a reputation list such as Spamhaus or SURBL that lists IPs, domains or URLs seen in abuse; on shared hosting one listing can block mail for every customer on the IP (Abuse Desks for Hosts & Registrars).
• Blur by default: showing possibly graphic media as a blurred placeholder the viewer chooses to reveal, as Discord does for teens and UK users (Terrorism, Violent Extremism & Graphic Content).
• Bot disclosure law (California): Business and Professions Code 17940 to 17943 (SB 1001, operative 1 July 2019), making it unlawful to use a bot to mislead a person about its artificial identity to sell goods or services or influence a vote; the duty falls on the bot's operator, not the platform (Spam, Bots & Inauthentic Behavior).
• Bot label: a platform-controlled marker that an account is automated, set by the operator or the platform and rendered by every client, tied to the human responsible (Spam, Bots & Inauthentic Behavior).
• BPO (business process outsourcing): the outsourcing firms that do most human review at the largest platforms, such as TaskUs and Teleperformance; the platform still owns the policy, the decision and the record (Moderation at Scale).
• Bridging: ranking that credits a note only for helpfulness that viewpoint doesn't explain, by regularizing intercepts more heavily than viewpoint factors so diverse raters must agree (Spam, Bots & Inauthentic Behavior).
• Bundling: NCMEC's feature letting large platforms combine related reports of one widespread incident, such as a viral meme, into a batched report that still names every user (Child Safety: Detection & Reporting).
• Christchurch Call: the commitments by governments and online service providers against terrorist and violent extremist content online, launched in May 2019 and now carried by the Christchurch Call Foundation (Terrorism, Violent Extremism & Graphic Content).
• CIB (coordinated inauthentic behavior): particularly sophisticated inauthentic behavior where false identities are central to the operation; enforced against the network, whatever it posted (Spam, Bots & Inauthentic Behavior).
• CIP (Content Incident Protocol): the IRF's first form, modeled on the response to Christchurch; retired in 2025 with the other old response types (Terrorism, Violent Extremism & Graphic Content).
• Claim dispute: an uploader's challenge to a fingerprinting claim (rights, an exception or misidentification), with a response window for the claimant that releases the claim if it lapses (Copyright at Scale).
• Classifier band: the thresholds for one model, category and action: above the upper one the system acts, between them a person reviews, below the lower one nothing happens; versioned data with an owner (Moderation at Scale).
• clientHold and serverHold: domain statuses that stop a name resolving, set by the registrar (clientHold) or the registry (serverHold); effective, and costly to every subdomain and mailbox on the name (Abuse Desks for Hosts & Registrars).
• Community Notes: crowd-written context attached to posts, shown only when people who usually disagree both rate a note helpful; launched on Twitter as Birdwatch and adopted by Meta and YouTube (Spam, Bots & Inauthentic Behavior).
• Compromised site: a real customer's site that was hacked, often an outdated CMS with a phishing kit dropped in; answered by blocking the paths, notifying the owner and setting a deadline (Abuse Desks for Hosts & Registrars).
• Content ID claim: a YouTube claim created when an upload matches a reference, with the owner's chosen action (block, monetize or track) per territory; it normally doesn't strike, unlike a removal request (Copyright at Scale).
• Content Safety API: Google's classifier for previously unseen abuse imagery that returns a priority, not a verdict; partners must review before acting (Child Safety: Detection & Reporting).
• Copyright strike: a penalty recorded against an uploader for a valid copyright removal, withdrawn by a valid counter-notice or retraction, counted toward termination (often three in a window) (Copyright at Scale).
• Counter-notice: a user's statement under penalty of perjury that material was removed by mistake or misidentification; on a valid one the platform restores the material 10 to 14 business days later unless the claimant sues (Copyright at Scale).
• Covered platform: under the TAKE IT DOWN Act, a public website or app that primarily provides a forum for user-generated content or regularly hosts non-consensual intimate depictions; nonprofits included, broadband and email excluded (Intimate Image Abuse).
• Cox v. Sony: the Supreme Court's 25 March 2026 holding that a provider is contributorily liable for copyright infringement only if it induced it or offered a service tailored to it; knowledge alone isn't enough (The Law of Hosting User Content).
• Crisis response protocol: a shared protocol for governments and companies to respond together to livestreamed terrorist content after an attack; GIFCT notifies the Christchurch Call and EU protocols on activation (Terrorism, Violent Extremism & Graphic Content).
• Cryptographic hash: a fingerprint such as MD5, SHA-1 or SHA-256 that matches only identical files; any change to the file changes it completely (Child Safety: Detection & Reporting).
• CSA image-generator: under the UK Crime and Policing Act 2026, a program, information or service made or adapted for creating child sexual abuse images; making, possessing or supplying one is an offence (prospective as of October 2026) (Child Safety: Detection & Reporting).
• CSA Regulation: the EU's proposed long-term framework against child sexual abuse online, in trilogue since December 2025 with detection still unresolved; no final text as of October 2026 (Child Safety: Detection & Reporting).
• CSAI Match: YouTube's video fingerprinting service for known abuse video, free to qualifying partners, who review matches before acting (Child Safety: Detection & Reporting).
• CSAM (child sexual abuse material): imagery of the sexual abuse of minors; every platform that accepts uploads, messages or streams from strangers will receive it, and US providers must report it on knowledge (Child Safety: Detection & Reporting).
• CSAM Scanning Tool: Cloudflare's free edge scanning that compares cached content against child-safety lists, blocks matches and emails the customer, leaving the reporting duty with the site owner (Abuse Desks for Hosts & Registrars).
• CSEA reporting duty (UK): the duty since 7 April 2026 for regulated user-to-user services to report detected, unreported child sexual exploitation and abuse content to the National Crime Agency, with three priority levels and clocks (Child Safety: Detection & Reporting).
• CyberTipline: NCMEC's reporting system for child sexual exploitation; providers report through its API, which mirrors the elements § 2258A lists (Child Safety: Detection & Reporting).
• CyberTipline preservation: the one-year duty to keep the reported contents and related data after a completed report, kept as a legal hold that every deletion and purge path respects (Child Safety: Detection & Reporting).
• DEFIANCE Act: S. 1837, which would add a federal civil action over intimate digital forgeries; passed the Senate in January 2026 and, as of October 2026, was held at the desk in the House (Intimate Image Abuse).
• Design defect claim: the theory in the social media addiction cases that features such as infinite scroll, autoplay, recommendations and notifications were negligently designed to be addictive, which puts product decisions into evidence (The Law of Hosting User Content).
• Designated agent: the person, position, department or service a platform registers with the US Copyright Office and publishes to receive DMCA notices; the designation expires after three years unless amended or resubmitted (Copyright at Scale).
• Digital Footprint Incident: an IRF activation type since 2025 for perpetrators or accomplices with a significant online presence who didn't produce media as part of the attack (Terrorism, Violent Extremism & Graphic Content).
• Digital forgery: an intimate depiction of an identifiable person created or altered with software or AI that a reasonable person would find indistinguishable from an authentic one (Intimate Image Abuse).
• Distributor liability: liability that attaches once a host is on notice; Zeran v. America Online (1997) held it is a form of publisher liability that Section 230(c)(1) bars (The Law of Hosting User Content).
• DJ Program: Twitch's 2024 licensing program with the major labels and Merlin for DJ streams, with licensing costs split between Twitch and DJs (Copyright at Scale).
• DMCA notice: a written notice of claimed infringement to the designated agent with the six elements of 512(c)(3)(A): signature, the work, the material and its location, contact, a good-faith statement and an accuracy statement under penalty of perjury (Copyright at Scale).
• DNS Abuse: in ICANN's contracts, malware, botnets, phishing, pharming, and spam only when used to deliver one of those (Abuse Desks for Hosts & Registrars).
• EDSA context: YouTube's educational, documentary, scientific or artistic exception for violent or graphic content, which requires the context to be in the video itself (Terrorism, Violent Extremism & Graphic Content).
• ENFORCE Act: a US bill to make penalties for AI-generated CSAM consistent with authentic CSAM; passed the Senate in December 2025 and, as of October 2026, is not law (Child Safety: Detection & Reporting).
• Error budget: the false-positive rate a platform accepts for automated action in a category, measured from daily samples of automated decisions; exceeding it narrows the band or returns the category to human review (Moderation at Scale).
• EU derogation: the temporary EU exception to ePrivacy confidentiality that lets messaging services voluntarily detect and report CSAM under conditions; Regulation 2021/1232 lapsed on 3 April 2026 and Regulation 2026/1881 revived it, excluding end-to-end encrypted content (Child Safety: Detection & Reporting).
• Evidence access log: an append-only record of every view, export and hold change on child-safety evidence, by named staff with a reason (Child Safety: Detection & Reporting).
• Exposed to terrorist content: an authority's decision under the TCO Regulation, for example after two or more final removal orders in 12 months, that obliges the provider to take specific measures of its choice and report on them (Terrorism, Violent Extremism & Graphic Content).
• Exposure cap: a per-reviewer daily limit on seconds of exposure to the most severe categories, enforced by the router and kept as data (Moderation at Scale).
• Fingerprinting: matching every upload against reference files supplied by rights holders, so copyright enforcement happens at upload time with an action the rights holder chose in advance (Copyright at Scale).
• Foreign interference: coordinated inauthentic behavior run by operators outside the country they target (Spam, Bots & Inauthentic Behavior).
• FOSTA: the 2018 amendment that added Section 230(e)(5), leaving civil sex-trafficking claims under 18 U.S.C. § 1595 and certain state prosecutions outside the immunity, while good-faith moderation keeps (c)(2)(A) protection (The Law of Hosting User Content).
• Fraudulent legal request: a forged court order, impersonated regulator or fake law enforcement request made to get speech removed or data disclosed; verified through a channel the platform found itself, and counted separately (The Law of Hosting User Content).
• GIFCT: the Global Internet Forum to Counter Terrorism, an independent industry organization that runs the shared hash database and the Incident Response Framework (Terrorism, Violent Extremism & Graphic Content).
• GIFCT labels: the entity, behavioral and incident labels on each shared hash, which let a member set policy per label rather than per list (Terrorism, Violent Extremism & Graphic Content).
• Good Samaritan clause (DSA Article 7): the EU rule that voluntary own-initiative investigations and measures against illegal content, done in good faith and diligently, don't cost a provider its liability exemptions (The Law of Hosting User Content).
• Grooming: an adult building contact with a child toward sexual abuse; detected through behavior and text and answered by restricting contact, preserving the conversation and reporting enticement (Child Safety: Detection & Reporting).
• Hash list: a list of hashes (or URLs) of material already judged, from one source under one access agreement; every match records the list and its version (Child Safety: Detection & Reporting).
• Hosted resource: a site, domain, IP address, server, bucket or DNS zone a provider can act on, recorded with who held it over time so a timestamped report resolves to the right customer (Abuse Desks for Hosts & Registrars).
• Hosting exemption (DSA Article 6): the EU rule that a host isn't liable for information it stores at a user's request unless it has actual knowledge (or, for damages, awareness of apparent illegality) and fails to remove it expeditiously (The Law of Hosting User Content).
• HSDB (hash-sharing database): GIFCT's shared store of hashes of known terrorist and violent extremist content, contributed and queried by members; a match prompts no automatic action on another member's platform (Terrorism, Violent Extremism & Graphic Content).
• Hybrid enforcement: an action where a reviewer decides an item violates policy and automated systems extend that decision to matching items, as Pinterest reports it; counted as automated, it overstates what machines decided alone (Moderation at Scale).
• ICP (information content provider): in Section 230, anyone responsible in whole or in part for creating or developing information; a platform is exposed for what it created or developed itself (The Law of Hosting User Content).
• Identical-copy sweep: the job that turns one removal request into removal of every exact copy across stores, a review lane for near matches, and the hashes added to the platform's own list (Intimate Image Abuse).
• Image Intercept: the IWF's lightweight hash-matching tool for small platforms, blocking a listed image before it is uploaded, stored or shared (Child Safety: Detection & Reporting).
• Image-based abuse scheme: Australia's Online Safety Act 2021 regime under which eSafety can require a service, poster or host to remove an intimate image within 24 hours (Intimate Image Abuse).
• Inauthentic behavior: in Meta's definition, deception performed by a network of inauthentic assets (Spam, Bots & Inauthentic Behavior).
• Inauthentic cluster: a network of accounts, communities, domains and content investigated and decided as one case, with evidence per member and any attribution recorded with its confidence (Spam, Bots & Inauthentic Behavior).
• Incident Advisory: an IRF activation type since 2025 for an extremely unclear information environment or an event that isn't obviously time-bound (Terrorism, Violent Extremism & Graphic Content).
• Incident mode: a platform's own response to one event: the incident's hashes loaded, tighter bands for them, a live queue, suppressed recommendations, and a recorded close (Terrorism, Violent Extremism & Graphic Content).
• Incidental and flagrant use: a rights contract's split between background or in-game music and radio, rebroadcast or pre-release use, each with its own action rather than one strike per notice (Copyright at Scale).
• Inducement: actively encouraging infringement through specific acts; after Cox v. Sony it is, with a service tailored to infringement, the route to contributory copyright liability (The Law of Hosting User Content).
• Informational report: a CyberTipline report NCMEC doesn't refer, because the information is insufficient or the imagery is viral and widely reported (Child Safety: Detection & Reporting).
• INHOPE: the international network of child sexual abuse hotlines (58 members in 53 countries by its own count) that passes reports across borders through its ICCAM tool (Child Safety: Detection & Reporting).
• Interactive blurring: review tooling where media starts blurred and the reviewer reveals only what the decision needs; shown to reduce emotional impact without losing accuracy or speed (Moderation at Scale).
• Interactive computer service: in Section 230, any information service, system or access software provider that enables computer access by multiple users to a computer server; the services the immunity protects (The Law of Hosting User Content).
• IRF (Incident Response Framework): GIFCT's protocol for attacks with an online dimension: alert members, share situational awareness and, for some activation types, open hash sharing for perpetrator content (Terrorism, Violent Extremism & Graphic Content).
• IWF (Internet Watch Foundation): the UK's hotline for online child sexual abuse imagery, a member-funded charity that runs lists, reporting portals and Report Remove and works with Ofcom (Child Safety: Detection & Reporting).
• IWF Hash List: the Internet Watch Foundation's licensed list of CSAM hashes, each item assessed by at least two people, with category metadata; about 3.2 million hashes at the end of 2025 (Child Safety: Detection & Reporting).
• IWF URL List: the IWF's licensed list of webpages with confirmed CSAM, updated twice a day, which members refresh at least every 24 hours (Child Safety: Detection & Reporting).
• JCCP (Judicial Council Coordination Proceeding): the California coordinated proceeding of social media addiction cases in Los Angeles Superior Court, where the first jury verdict against Meta and YouTube was returned on 25 March 2026 (The Law of Hosting User Content).
• Known identical copies: the copies a covered platform must make reasonable efforts to find and remove within the same 48 hours; exact hash matches, with near matches reviewed (Intimate Image Abuse).
• Lantern: the Tech Coalition's cross-platform signal-sharing program for child sexual exploitation (41 companies as of October 2026), where each participant reviews incoming hashes, URLs and incident signals before acting (Child Safety: Detection & Reporting).
• Legal clock: the deadline a regime attaches to knowledge, such as one hour for an EU terrorist content order, 48 hours for a TAKE IT DOWN Act request, or 10 to 14 business days for a DMCA restore; computed once from a versioned rule and stored on the notice (The Law of Hosting User Content).
• Legal notice: any notice or order that carries a legal clock, whatever law it comes under (DMCA takedowns and counters, TAKE IT DOWN Act requests, DSA notices, TCO orders, court and regulator orders), kept as one legal_notices row with its named items (The Law of Hosting User Content).
• M3AAWG hosting practices: the M3AAWG and Internet Infrastructure Coalition Anti-Abuse Best Common Practices for Hosting and Cloud Service Providers (2015), the written baseline for abuse handling and priority (Abuse Desks for Hosts & Registrars).
• Malicious registration: a domain or account created by the criminal for the attack, often in bulk; answered by suspending it outright and looking for related registrations (Abuse Desks for Hosts & Registrars).
• MDL 3047: the federal multidistrict litigation over social media addiction in the Northern District of California; as of October 2026 it has no bellwether verdict (The Law of Hosting User Content).
• Mitigation layers: the levels at which an infrastructure provider can act (content, site, account, domain, DNS, network), each wider one stopping more abuse and hurting more bystanders (Abuse Desks for Hosts & Registrars).
• Music Reporting Process: Twitch's process, separate from the DMCA, that lets participating rights holders report music in streams and VODs under activity-specific penalty ladders (Copyright at Scale).
• Narrow and scaled allowances: Meta's two forms of newsworthiness allowance: narrow covers one item, scaled covers a phrase or broader class and should expire and be reviewed (Terrorism, Violent Extremism & Graphic Content).
• NCMEC: the National Center for Missing & Exploited Children, the US clearinghouse that runs the CyberTipline, hash lists and Take It Down (Child Safety: Detection & Reporting).
• NCMEC NGO hash list: NCMEC's list of more than five million hash values of confirmed, triple-vetted CSAM, available to interested technology companies (Child Safety: Detection & Reporting).
• Needs review label: Bluesky's temporary label applied automatically to accounts that look like large-scale spam, limiting impact while moderators decide (Spam, Bots & Inauthentic Behavior).
• Newsworthiness allowance: a documented exception that keeps up content the rules would remove because the public interest outweighs the risk of harm; granted by a policy role, not a first-line reviewer (Terrorism, Violent Extremism & Graphic Content).
• No general monitoring (DSA Article 8): the EU rule that providers can't be given a general obligation to monitor what they transmit or store, or to actively seek facts indicating illegal activity (The Law of Hosting User Content).
• No-view actioning: removing content, quarantining the asset and restricting the account on an exact match to a vetted list without anyone opening the file (Child Safety: Detection & Reporting).
• Non-Photographic Imagery URL List: the IWF list of pages with computer-generated, drawn or animated child sexual abuse imagery (Child Safety: Detection & Reporting).
• Note intercept: a Community Notes note's score: the helpfulness left after the model's viewpoint factors explain what they can; at least 0.40 counted as helpful in the published algorithm (Spam, Bots & Inauthentic Behavior).
• Notice and stay down: the Article 17 duty that after a valid notice the same work shouldn't come back, implemented by making the notified work a fingerprint reference (Copyright at Scale).
• Nudify app: an app or site that generates intimate images from any photo of a real person; platforms and prosecutors act against its ads, links and operators (Intimate Image Abuse).
• Null route: dropping all traffic to an IP address at the network layer; it stops a live attack and takes down every site on a shared IP (Abuse Desks for Hosts & Registrars).
• OCSSP (online content-sharing service provider): under Article 17 of the DSM Directive, a service whose main purpose includes storing and giving access to a large amount of user-uploaded protected works that it organises and promotes for profit (Copyright at Scale).
• PAT (Private Access Token): a Privacy Pass token issued after the device vendor attests a real device, letting a site skip a challenge while learning only that a valid token was presented (Spam, Bots & Inauthentic Behavior).
• PDQ and TMK+PDQF: Meta's open-source perceptual hashes for photos (PDQ) and video (TMK+PDQF), released in 2019 so hash-sharing systems can interoperate (Child Safety: Detection & Reporting).
• PERCI: Europol's platform for processing EU removal orders and referrals, launched in July 2023 (Terrorism, Violent Extremism & Graphic Content).
• Perpetrator Content Incident: an IRF activation type since 2025 for online, perpetrator-produced content directly related to a terrorist or violent extremist event (Terrorism, Violent Extremism & Graphic Content).
• Pex: a music and video identification vendor acquired by Vobile in April 2025 (Copyright at Scale).
• PhotoDNA: Microsoft's image hashing for matching known illegal images, offered free to qualified organizations and as a cloud service (Child Safety: Detection & Reporting).
• Prevalence: the estimated share of views of violating content among all views, measured by sampling views and labeling them; the only number that captures what moderation missed (Moderation at Scale).
• Proof of work: a challenge that makes the client spend CPU time per request, trivial for one sign-up and expensive for a hundred thousand; it separates cheap from expensive, not people from machines (Spam, Bots & Inauthentic Behavior).
• RDAP: the Registration Data Access Protocol, the definitive source for generic top-level domain registration data since 28 January 2025, with an abuse entity role naming who handles abuse (Abuse Desks for Hosts & Registrars).
• Red-flag knowledge: awareness of facts or circumstances from which infringement is apparent; under the DMCA it ends the 512(c) safe harbor like actual knowledge unless the platform removes the material expeditiously (The Law of Hosting User Content).
• Reduced distribution: keeping an item visible on direct link or profile while excluding it from recommendations, search suggestions and trending (Terrorism, Violent Extremism & Graphic Content).
• Referral (CyberTipline): a report NCMEC sends to a specific law enforcement agency because the company gave enough information (user details, imagery, a possible location) (Child Safety: Detection & Reporting).
• Referral (terrorist content): content an authority or Europol flags for a provider's voluntary assessment against its own terms, with no legal clock (Terrorism, Violent Extremism & Graphic Content).
• Removal order: an order from a competent EU authority under the TCO Regulation to remove or disable terrorist content in all Member States within one hour of receipt (Terrorism, Violent Extremism & Graphic Content).
• Repeat infringer policy: the reasonably implemented policy, required for every DMCA safe harbor, of terminating repeat infringers in appropriate circumstances; usually strikes with expiry (Copyright at Scale).
• REPORT Act: the 2024 law that added sex trafficking of minors and enticement to the reporting duty, extended preservation from 90 days to one year, and raised fines (Child Safety: Detection & Reporting).
• Report Remove: the IWF's service letting young people in the UK report their own images for removal (Child Safety: Detection & Reporting).
• Rights reference (reference file): a rights holder's reference file with its owner, territories, per-territory policy and exclusivity attestation; under Article 17 it is the relevant and necessary information (Copyright at Scale).
• Safer: Thorn's commercial child-safety product: perceptual hashing for images and video, classifiers for novel CSAM, and a text classifier for exploitation and sextortion (Child Safety: Detection & Reporting).
• Scola v. Facebook: the California moderator class action whose settlement got final approval on 14 July 2021, with a $52 million fund and terms (muted grayscale video, counseling, vendor duties) as reported, the best-documented reviewer protections (Moderation at Scale).
• Section 230(c)(1): the US rule that no provider or user of an interactive computer service is treated as the publisher or speaker of information another information content provider supplied; unlike the other regimes, notice doesn't remove it (The Law of Hosting User Content).
• Section 230(c)(2): the companion immunity for good-faith actions to restrict access to objectionable material, so moderating some content doesn't make a platform the publisher of the rest (The Law of Hosting User Content).
• Section 512 safe harbors: the DMCA's four copyright safe harbors by function: transitory communications (512(a)), caching (512(b)), information stored at users' direction (512(c)) and information location tools (512(d)) (The Law of Hosting User Content).
• Sextortion: coercing a victim, often a minor, with threats to share sexual images, including financial sextortion run through payment handles and gift cards (Child Safety: Detection & Reporting).
• Shadow mode: running a candidate model or vendor on live traffic with its outputs logged and ignored, then comparing against labeled decisions before letting it act (Moderation at Scale).
• Signal vendor: a supplier that returns scores or matches (Hive, Community Sift, Safer, Google's Content Safety API); changing one changes a signal (Moderation at Scale).
• Skewing (OAT-016): OWASP's automated-threat name for repeated clicks, requests or submissions intended to alter a metric such as likes, votes or poll results (Spam, Bots & Inauthentic Behavior).
• Small-service regime: Article 17's lighter conditions for services in the EU under three years with turnover below EUR 10 million; stay-down applies once monthly visitors exceed 5 million (Copyright at Scale).
• SMS pumping: fraud that abuses a phone number field to send one-time codes to blocks of numbers on a carrier that shares the termination revenue; also called artificially inflated traffic (Spam, Bots & Inauthentic Behavior).
• STM (standard technical measures): identification and protection measures developed by broad multi-industry consensus and available on reasonable terms, which DMCA safe harbors require platforms to accommodate (Copyright at Scale).
• StopNCII: SWGfL's service where adults hash their intimate images on their own device and partner platforms match and remove copies under their policies (Intimate Image Abuse).
• Substantial compliance: the DMCA test for notices: one that doesn't substantially comply can't show knowledge, but one that identifies the work, the material and the contact obliges the platform to try to get a compliant notice (Copyright at Scale).
• Sunset Section 230 Act: S. 3546, a December 2025 bill that would repeal Section 230 two years after enactment; as of October 2026 its only action is referral to committee (The Law of Hosting User Content).
• Sunset To Reform Section 230 Act: H.R. 6746, a December 2025 bill that would end Section 230's effect after 31 December 2026; as of October 2026 it has only been referred to committee (The Law of Hosting User Content).
• Take It Down: NCMEC's service for images taken before the person was 18: the hash goes to NCMEC's list and participating platforms scan public or unencrypted services against it (Intimate Image Abuse).
• TAKE IT DOWN Act (TIDA): the 2025 US law that criminalizes publishing non-consensual intimate depictions and digital forgeries and, since 19 May 2026, requires covered platforms to remove reported images within 48 hours of a valid request (Intimate Image Abuse).
• TCO Regulation: the EU's Terrorist Content Online Regulation (EU) 2021/784, applying since June 2022 to hosting providers that disseminate to the public, with one-hour removal orders (Terrorism, Violent Extremism & Graphic Content).
• Transfer lock: a status that stops a domain moving to another registrar, used with a suspension to prevent evasion (Abuse Desks for Hosts & Registrars).
• Trusted reporter: a volume reporter, such as a blocklist operator or CERT, given a priority lane at an abuse desk; the infrastructure counterpart of a trusted flagger (Abuse Desks for Hosts & Registrars).
• UK-linked: in the UK reporting duty, content with evidence linking it to the UK, such as where it was uploaded, the offender's nationality or location, or a suspected victim's location (Child Safety: Detection & Reporting).
• Valid removal request: a TAKE IT DOWN Act request with a signature, information to locate the depiction, a good-faith statement that it isn't consensual, and contact information; the 48-hour clock runs from it (Intimate Image Abuse).
• Viewed flag (fileViewedByEsp): the CyberTipline field recording whether a person at the platform viewed a file before reporting; after United States v. Wilson it changes what law enforcement can do without a warrant (Child Safety: Detection & Reporting).
• VOD muting: silencing archived broadcast audio where third-party music is detected; Twitch began it in 2014 in 30-minute blocks (Copyright at Scale).
• Vote fuzzing: altering visible vote totals so manipulators can't tell whether their votes counted; Reddit admins did it as an anti-spam measure (Spam, Bots & Inauthentic Behavior).
• Vote ring: a group of accounts that coordinate votes to push content up or down; detected as a cluster of voters, not single votes (Spam, Bots & Inauthentic Behavior).
• VVR (violative view rate): YouTube's prevalence measure: the share of views on videos that violate its Community Guidelines, estimated by sending a sample of videos to reviewers (Moderation at Scale).
• Warning screen (interstitial): a full overlay naming the kind of content that a viewer clicks through, used for graphic material kept up for news, documentary or awareness (Terrorism, Violent Extremism & Graphic Content).
• Workflow vendor: a supplier that holds the queues, policies, decisions and regulatory filings (Checkstep, Cinder); changing one is a migration of the moderation system of record (Moderation at Scale).
• XARF: a set of JSON schemas for abuse reports (phishing, malware, copyright, DDoS, child exploitation and more) developed by Abusix, carried by email, API or stream (Abuse Desks for Hosts & Registrars).

The Rules#

• 2703(d) order: a US court order under the Stored Communications Act, issued on "specific and articulable facts," that compels non-content records such as login history and the participant list of a chat (Law Enforcement Requests & User Data).
• Access laws: US state laws that require age verification for all users and parental consent for minors' accounts, most of them challenged in court by NetChoice and CCIA (Minors, Age Assurance & US State Law).
• Action plan: the remedy a DSA fine comes with, a project with deadlines measured in working days and periodic penalties behind it, such as the plans the Commission required from X, Temu and AliExpress (Regulators & Enforcement).
• Actor type: who made a moderation decision (volunteer, staff, or automation acting for either), recorded on the decision so a report can show who moderates the platform (Transparency Reporting).
• Ad repository (Article 39): the searchable repository of ads shown, who paid and whom they targeted, which very large platforms must keep under the DSA and which was one of the three grounds for X's fine (Regulators & Enforcement).
• Addictive feed: California SB 976's term for a personalized feed that requires parental consent for minors, a provision partly in effect as of October 2026 (Minors, Age Assurance & US State Law).
• Age assurance: the methods a platform uses to establish a user's age (self-declaration, platform signals, inference, facial estimation, document checks), picked per feature and layered so the strongest check goes to the contested cases (Minors, Age Assurance & US State Law).
• Age assurance funnel: users prompted to verify their age, those who completed, those who passed, and the method used, which Discord publishes monthly and which every age check row makes a group-by (Minors, Age Assurance & US State Law).
• Age band: the cached band on the account (accounts.age_band, from under 13 upward), derived lowest-wins from current age signals and re-evaluated at each sensitive action, with unknown treated as youngest (Minors, Age Assurance & US State Law).
• Age check: one attempt to establish a user's age, from prompt to result, recorded whether or not the user finishes, with its method, vendor reference and the time its evidence was deleted (Minors, Age Assurance & US State Law).
• Age inference: placing a user's age from signals such as tenure, device and activity data, which lets a platform verify only the people it can't already place (Minors, Age Assurance & US State Law).
• Age restriction (account): a record that an account is below a minimum age or limited by age in a jurisdiction; it isn't a penalty, doesn't count toward standing, and ends on proof of age or a birthday (Minors, Age Assurance & US State Law).
• Age-Appropriate Design Code (California AB 2273): California's design code for minors, whose coverage, age estimation and data-use provisions the Ninth Circuit let proceed in March 2026 while its dark patterns provision stayed enjoined (Minors, Age Assurance & US State Law).
• Age-Restricted Material Codes: Australian industry codes registered by eSafety, three in effect from December 27, 2025 and six more, including social media and messaging features, from March 9, 2026 (Regulators & Enforcement).
• Age-restricted social media platform: a platform eSafety treats as covered by Australia's under-16 minimum age, a list that includes Facebook, Instagram, Snapchat, Threads, TikTok, Twitch, X, YouTube, Kick and Reddit "among others" (Regulators & Enforcement).
• Annotation: a recorded explanation of why a reported number moved (a policy change, new tooling, an error), stored with the period rather than written into prose at publication (Transparency Reporting).
• App-store laws: US state laws that move the age check to Apple and Google rather than each app (Minors, Age Assurance & US State Law).
• Appeals Centre Europe: an out-of-court dispute body certified in Ireland that received 24,000 disputes from April 2025 to March 2026 and disagreed with the platform in 59% of the decisions where it could review the content (Regulators & Enforcement).
• Article 15 report: the DSA transparency report every intermediary service publishes at least yearly, covering authority orders, notices, own-initiative moderation, complaints and the accuracy of automated tools; micro and small enterprises are exempt unless designated very large (Transparency Reporting).
• Article 16 notice: a notice of illegal content under the DSA's notice and action duty, which can come from a non-user through a web form as well as from a member, and which a platform counts by type and handling time (Regulators & Enforcement).
• Article 24 report: the DSA additions for online platforms: out-of-court dispute counts and the share implemented, Article 23 suspensions, and average monthly active recipients in the EU at least every six months (Transparency Reporting).
• Article 42 report: the six-monthly DSA report very large platforms publish, adding moderation staff and accuracy indicators per official language and recipients per member state (Transparency Reporting).
• Average monthly active recipients: the count of EU users every online platform must publish at least every six months under DSA Article 24(2), and the number the Commission designates very large platforms from (Regulators & Enforcement).
• Basic Online Safety Expectations: Australia's ministerial determination of what social media, messaging, gaming and other services should do; not enforceable on their own, but eSafety can compel reports on how a provider meets them (Regulators & Enforcement).
• Basic subscriber information: name, address, session times, length and type of service, assigned IP address and means of payment, the least protected tier of the Stored Communications Act, compellable by subpoena (Law Enforcement Requests & User Data).
• Business disruption measures: court orders Ofcom can seek in the most serious cases that require payment providers or advertisers to withdraw, or internet service providers to block access in the UK (Regulators & Enforcement).
• Categorisation register: Ofcom's list of categorised services, published July 10, 2026; Category 1 carries user empowerment, journalistic content and fraudulent advertising duties, Category 2A adds risk assessment record-keeping, and Categories 1, 2A and 2B all publish annual transparency reports (Regulators & Enforcement).
• Civil subpoena: a private party's demand for records, which can reach non-content records but never content, because the Stored Communications Act has no civil discovery exception (Law Enforcement Requests & User Data).
• CLOUD Act: the 2018 US law that requires providers to produce data in their control wherever it is stored, and that lets the US conclude executive agreements under which partners' authorities send orders directly to US providers (Law Enforcement Requests & User Data).
• CLOUD Act executive agreement: a bilateral agreement, such as the US-UK (in force October 3, 2022) and US-Australia (January 31, 2024) agreements, under which a partner's authorities order data directly from US providers for serious crime (Law Enforcement Requests & User Data).
• Codes of practice (Online Safety Act): Ofcom's measures that count as compliance if followed; a provider may choose other measures but must show they work, and the codes are amended by statement (Regulators & Enforcement).
• Coimisiún na Meán: Ireland's Digital Services Coordinator, the DSA regulator for providers established in Ireland, which certifies dispute bodies and awards trusted flagger status (Regulators & Enforcement).
• Commitments: binding undertakings a platform makes to the Commission under the DSA, as AliExpress did on ad and recommender transparency and TikTok did on its ad repository (Regulators & Enforcement).
• Contents: posts, comments, messages, chats, votes and uploaded files, the most protected Stored Communications Act tier, which in practice needs a warrant (Law Enforcement Requests & User Data).
• COPPA age-check policy statement: the FTC's February 25, 2026 statement that it won't enforce COPPA against general or mixed audience operators that collect data solely to determine age, if they limit its use, retention and sharing and use a reasonably accurate method (Minors, Age Assurance & US State Law).
• Default decision: a dispute body's ruling for the user because the platform didn't provide the content; Appeals Centre Europe issued more than 7,000, mostly about account suspensions (Regulators & Enforcement).
• Defaults by age: the settings that change for minors (who can contact, who can find and see, what they see, what they can do, how they're kept), each with a regulatory anchor and stored as rule data (Minors, Age Assurance & US State Law).
• Design laws: US state laws that limit "addictive" feeds, notifications and time of use, or impose age-appropriate design codes (Minors, Age Assurance & US State Law).
• Design liability: claims that a feature's design (infinite scroll, autoplay, notifications), not the content, harmed a young user, as in the March 2026 Los Angeles verdict against Meta and YouTube (Minors, Age Assurance & US State Law).
• Designated addressee: the establishment or legal representative a provider offering services in the EU must designate and register to receive e-evidence orders (Law Enforcement Requests & User Data).
• Designation: a regulator placing a service in a tier, such as the Commission naming a very large online platform; under the DSA the tier's duties apply four months after notification, so a designation is a product deadline (Regulators & Enforcement).
• Digital Services Act (DSA): the EU regulation that stacks duties by tier (intermediary, hosting service, online platform, very large online platform or search engine) and lets regulators fine up to 6% of worldwide annual turnover (Regulators & Enforcement).
• Digital Services Coordinator: the national DSA regulator in each member state, which enforces against everyone below the very large tier, certifies dispute bodies and awards trusted flagger status (Regulators & Enforcement).
• DPIA (data protection impact assessment): the risk assessment the ICO found Imgur had never done, one of the three failures behind its £247,590 fine (Minors, Age Assurance & US State Law).
• E-Evidence Regulation: Regulation (EU) 2023/1543, applying from August 18, 2026, which lets a judicial authority send production and preservation orders directly to a provider's representative in another member state (Law Enforcement Requests & User Data).
• Emergency disclosure: a provider's discretionary disclosure of contents or records without process when it believes in good faith that an emergency involving danger of death or serious physical injury requires it, the request type most targeted by fraud (Law Enforcement Requests & User Data).
• Enforcement ledger: a dated record of every regulator contact, request, finding, commitment and deadline with the people responsible, kept for the platform's own matters (Regulators & Enforcement).
• eSafety Commissioner: Australia's online safety regulator under the Online Safety Act 2021, which administers the Basic Online Safety Expectations, industry codes and the social media minimum age (Regulators & Enforcement).
• European Preservation Order: an e-evidence order to hold data for 60 days, extendable once by 30, and longer if a production request follows (Law Enforcement Requests & User Data).
• European Production Order: an e-evidence order for subscriber, identifying, traffic or content data, answered within 10 days, or within eight hours in an emergency (Law Enforcement Requests & User Data).
• Facial age estimation: estimating age from a selfie, on the device in Discord's design, whose error is largest near the threshold, so a buffer zone around each age line goes to a stronger method (Minors, Age Assurance & US State Law).
• Family supervision: opt-in, reversible guardian tools such as Discord's Family Center that show a teen's activity (friends, servers, who they messaged) but not message content (Minors, Age Assurance & US State Law).
• Feature age rule: a dated row mapping a feature, jurisdiction and age band to a mode with parameters and a cited source, so an injunction is a new row rather than a deploy (Minors, Age Assurance & US State Law).
• Free Speech Coalition v. Paxton: the June 2025 Supreme Court decision that upheld Texas's age verification law for sites where more than one-third of the material is sexual material harmful to minors, applying intermediate scrutiny (Minors, Age Assurance & US State Law).
• Geo restriction: a row that withholds an item, account, community, feature or the whole service in a country, checked at serving time against the viewer's location and answered with HTTP 451 and the legal basis (Regulators & Enforcement).
• Geoblock: blocking a service or item for viewers in one country, a control a regulator expects the platform to prove works across account country, IP, every host name, embeds and APIs (Regulators & Enforcement).
• Guardian link: a consented, reversible connection between a guardian and a teen account that records what the teen agreed to share and who can end it, kept separate from parental consent (Minors, Age Assurance & US State Law).
• Harmonised templates: the CSV and XLSX templates fixed by Commission Implementing Regulation (EU) 2024/2835 that DSA transparency reports must use from July 1, 2025, with quantitative and qualitative parts (Transparency Reporting).
• HTTP 451 (Unavailable For Legal Reasons): the status code for content withheld by legal demand, whose body should explain the demand and whose Link header with rel="blocked-by" names the blocking entity (Regulators & Enforcement).
• ICO (Information Commissioner's Office): the UK data protection regulator, which fined Imgur's owner over children's data and said that exiting the UK doesn't avoid responsibility (Regulators & Enforcement).
• Independent audit (Article 37): the audit a very large platform must undergo at least once a year by an independent organization, with an implementation report within one month of an audit that isn't positive (Regulators & Enforcement).
• KIDS Act (Kids Internet and Digital Safety Act): H.R. 7757, passed by the House on June 29, 2026, which would default platforms to their most protective settings for minors and preempt conflicting state laws, while saying outright that it imposes no duty of care (KOSA's core) (Minors, Age Assurance & US State Law).
• Legal ground versus terms ground: whether a moderation action rested on the law or on the platform's own terms, a per-action column the DSA templates require (Transparency Reporting).
• Legal representative (Article 13): the person a provider with no EU establishment must designate in a member state, whom authorities can address instead of the provider and who can be held liable (Regulators & Enforcement).
• Legal request: one request for user data from an authority or private party, with its process type, verification, deadlines and status, plus the identifiers it named and exactly what was produced (Law Enforcement Requests & User Data).
• Minimum age: a legal floor on account holders, such as Australia's under-16 rule from December 10, 2025, which in engineering terms means finding existing underage accounts, restricting with an appeal, stopping new sign-ups and keeping evidence of reasonable steps (Minors, Age Assurance & US State Law).
• Mutual legal assistance (MLAT): the treaty route in which a foreign government asks the US government, which obtains US process for the platform to answer (Law Enforcement Requests & User Data).
• National security letter: a national security demand that, like a FISA order, sits on the same ladder and may be reported only in ranges, which is why Pinterest reports such requests as "0 to 249" (Law Enforcement Requests & User Data).
• Non-periodic notice: a one-off eSafety notice requiring a provider to report how it meets the Basic Online Safety Expectations on a topic; X was penalized A$650,000 for a response that didn't follow the specified manner and form (Regulators & Enforcement).
• Nondisclosure order (2705(b)): a court order forbidding a provider to notify anyone of a request, for a period the court sets; a request that merely asks for confidentiality isn't one (Law Enforcement Requests & User Data).
• Notifier kind: who sent a report (a member, a non-user, a trusted flagger or an authority), recorded on the report so intake can set the priority lane and the Article 15 report can count trusted flaggers' notices (Regulators & Enforcement).
• Ofcom: the UK regulator under the Online Safety Act, which sets codes, keeps the categorisation register, issues transparency notices and fines, and can seek business disruption measures (Regulators & Enforcement).
• Online Safety Act (UK): the law that applies to user-to-user and search services with links to the UK wherever the provider is based, with fines up to £18 million or 10% of qualifying worldwide revenue (Regulators & Enforcement).
• Out-of-court dispute settlement (Article 21): a DSA step between the internal appeal and a court, in which a certified body hears a user's dispute about a moderation decision within 90 days; decisions aren't binding, and the platform pays the fees if the user wins (Regulators & Enforcement).
• Pen register and trap and trace order: a US order for ongoing collection of future non-content data such as IP logs and headers for a defined period, never content (Law Enforcement Requests & User Data).
• Periodic penalty payment: a DSA penalty of up to 5% of average daily income or turnover per day (Article 76), which stands behind each action plan's deadlines (Regulators & Enforcement).
• Preliminary findings: the Commission's preliminary view that a platform breached the DSA, as against TikTok's addictive design in February 2026 (Regulators & Enforcement).
• Preservation request (2703(f)): a government request that obliges a provider to preserve existing records for 90 days, extendable by another 90, pending process; it copies what exists and discloses nothing (Law Enforcement Requests & User Data).
• Process type: the legal kind of a request (subpoena, 2703(d) order, warrant, pen register order, preservation, emergency, MLAT, CLOUD Act order, European order, civil subpoena), which decides the data categories that may be produced (Law Enforcement Requests & User Data).
• Production: exactly what left the platform for a request: the approved categories and date range, packaged with a SHA-256 manifest, approved and produced by two different people (Law Enforcement Requests & User Data).
• Reach band: Pinterest's measure of how many unique viewers saw content before it was deactivated, an exposure figure beside removal counts (Transparency Reporting).
• Regime duty: a dated rule row recording what a regime (DSA, Online Safety Act, eSafety and others) requires of a service tier, with its source, so "what must this service do in this country today?" is a query (Regulators & Enforcement).
• Regulator request: an information request, transparency notice or inquiry from a regulator, handled like legal process with an owner, a stored deadline and a record of exactly what was sent (Regulators & Enforcement).
• Report actionability: the share of user reports that led to action, a statement about report quality (Reddit's was 6.9% in the second half of 2025) (Transparency Reporting).
• Researcher data access (Article 40): the DSA duty on very large platforms to give vetted researchers data and not block access to publicly accessible data through their terms (Regulators & Enforcement).
• Section 77 transparency report: the annual report each UK categorised service publishes with the information Ofcom specifies in a per-service transparency notice, so it isn't a fixed template (Transparency Reporting).
• Social media minimum age: Australia's rule that, from December 10, 2025, age-restricted social media platforms take reasonable steps to keep under-16s from creating or keeping accounts, with court penalties up to 300,000 penalty units (A$109.2 million per eSafety) (Regulators & Enforcement).
• Statement of reasons (Article 17): the DSA explanation a user receives for a moderation decision, including its legal ground and territorial scope, which is also submitted to the Transparency Database; the book stores it on the decision row (Regulators & Enforcement).
• Stop Hiding Hate Act: New York's law requiring social media companies with more than $100 million in annual gross revenue to file their terms of service and moderation data with the Attorney General twice a year (Transparency Reporting).
• Stored Communications Act (SCA): the US statute (18 U.S.C. 2701 to 2713) that governs what a public provider may disclose and what the government can compel, as a ladder from subscriber data to content (Law Enforcement Requests & User Data).
• Superseded report: a corrected transparency report published as a new row that replaces the old one while both stay public; a published value is never edited (Transparency Reporting).
• Systemic risk assessment (Article 34): the very large platform duty to assess risks from the service and its design, including illegal content, fundamental rights, elections, minors and well-being, followed by mitigation under Article 35 (Regulators & Enforcement).
• Teen-by-default: treating every user whose age is unknown as a teen and letting verified adults opt out, as Discord did globally from September 23, 2026, after delaying its March plan (Minors, Age Assurance & US State Law).
• Transparency Database: the Commission's database that receives each statement of reasons a platform issues, a per-decision stream that should reconcile with the Article 15 counts (Transparency Reporting).
• Transparency notice (UK): Ofcom's annual notice telling a categorised service what to publish in its transparency report; the first draft notices went out from November 25, 2026 (Regulators & Enforcement).
• Transparency report: a platform's public account of its own moderation (removals, detection, appeals, child safety reports and government requests), voluntary for a decade and now also regulated in the EU, UK, Australia and New York (Transparency Reporting).
• Transparency report value: one published number stored with its definition version, dimensions, unit and snapshot, so any figure can be traced to the query that produced it (Transparency Reporting).
• Trusted flagger (Article 22): an entity a Digital Services Coordinator has recognized for expertise and accuracy, whose notices in its field must get priority; the platform syncs the Commission's list and reports flaggers who send poor notices (Regulators & Enforcement).
• Unit of enforcement: what a report counts (a post, a distinct image, a Pin, an account, an action), the choice that makes transparency reports incomparable and that each report should state (Transparency Reporting).
• User notice: telling a user their data was requested before production, the default at every platform surveyed, suppressed only by a court order, an emergency, child safety or a hijacked account, and scheduled for when a prohibition ends (Law Enforcement Requests & User Data).
• Verified adult: an account whose adult age has been established by a check, the status Discord's teen-by-default requires to unblur sensitive content, enter age-restricted spaces or (in its February 2026 announcement) speak on a Stage (Minors, Age Assurance & US State Law).
• Very large online platform (VLOP): a platform with at least 45 million average monthly active recipients in the EU, designated by the Commission and supervised by it directly; Reddit was designated on August 31, 2026 (Regulators & Enforcement).
• Voluntary disclosure (2702): the Stored Communications Act's general bar on a public provider divulging contents, and records to government, except under listed exceptions such as emergencies or user consent (Law Enforcement Requests & User Data).
• Withdrawal: leaving a market rather than complying, as Imgur did for the UK; past processing remains enforceable after the exit (Regulators & Enforcement).
• Withholding: blocking one item in one country on a legal demand while keeping it up elsewhere, the narrowest of a platform's four geographic responses (Regulators & Enforcement).

Running the Platform#

• 410 Gone: the status a removed item's URL answers when it was removed for policy, deleted by its author or purged, with a coarse public reason in the body (Migrations, Purges & Shutdowns).
• Acceptance rate: the share of booking requests hosts accept, with median response time and the share that expire; a request-to-book marketplace with a low rate fails guests even when search converts (Operating Metrics).
• Active supply: the creators, streamers, sellers, hosts or listings active in a period, the leading indicator for creator and marketplace businesses (Operating Metrics).
• Activity fact: one row per identity (account or anonymous identifier) per day it was active, from which DAU, DAUq-style counts, MAU and the DSA's EU recipients are all different queries (Operating Metrics).
• AI answers: a model answering questions over the platform's own corpus, such as Reddit Answers in core search, which must cite threads, respect removals and visibility, and be moderated as the platform's own speech (AI on User-Generated Platforms).
• AI crawler blocking: refusing unlicensed AI collection at the edge; Cloudflare made it its default on July 1, 2025 unless crawlers pay, so crawler policy became a CDN setting (AI on User-Generated Platforms).
• AI label: a mark on content saying it was AI-generated or altered, set by the uploader's declaration, by detection, from provenance or after review; detected and reviewed labels are moderation decisions with notice and appeal (AI on User-Generated Platforms).
• AI training preference: a user's dated choice to allow or opt out of third-party AI use of their content, read by licensed exports at export time and propagated to partners when it changes (AI on User-Generated Platforms).
• Article 50 (EU AI Act): the AI Act's transparency article, applying since August 2, 2026, that requires deepfake and public-interest text disclosure by deployers, machine-readable marking by providers, and chatbot disclosure (AI on User-Generated Platforms).
• Booked and served clocks: the two dates on every marketplace transaction, when demand happened (what gross booking value counts) and when the stay or trip happened (what revenue counts) (Operating Metrics).
• Cell: one of many smaller search clusters, Discord's replacement for two huge Elasticsearch clusters, so the biggest communities can get dedicated indices (Platform Architecture).
• Chatbot disclosure: the AI Act duty (Article 50(1)) to tell people they're talking to an AI unless it's obvious, which reaches AI assistants and AI accounts that converse with users (AI on User-Generated Platforms).
• Classifier policy: an effective-dated row holding the threshold, target precision and measured precision for one policy category and automated action, with the evaluation that justified it and its approver (AI on User-Generated Platforms).
• Community health: active communities, communities with an active moderator, and moderator actions per moderator; Reddit banned 605,570 communities in the second half of 2025, mostly for being unmoderated or spam (Operating Metrics).
• Community move: relocating a community that outgrew its shard or cell through copy, catch-up from the event log, a brief write pause and cutover, keeping the old copy until reconciliation passes (Platform Architecture).
• Community placement: a row with history recording where a community's partitioned data lives, on a shared or dedicated tier, which the router reads and a legal or residency audit can query by date (Platform Architecture).
• Concentration: how much content comes from the top 1% of creators and how much consumption goes to it, both a dependency risk and a moderation risk (Operating Metrics).
• Content tombstone: the row that outlives a removed item, holding its id, kind, community, coarse public reason, date and status code but no content, which the serving path and CDN read to answer 404, 410 or 451 (Migrations, Purges & Shutdowns).
• Country exit: a shutdown with a border, as Imgur's UK block and Twitch's Korea closure were: scope set as a service-wide geo restriction, rights requests kept open, money settled, embeds and APIs covered and records kept (Migrations, Purges & Shutdowns).
• Creators as a share of actives: accounts that posted, commented, uploaded, streamed or listed divided by actives, where a small change is a large change in supply (Operating Metrics).
• Data export: a user's re-authenticated, resumable and expiring copy of their data, with originals and history, recording what was excluded, such as others' content or data under a nondisclosure order (Migrations, Purges & Shutdowns).
• Data portability (GDPR Article 20): the right to personal data the user provided, processed on consent or contract by automated means, in a structured, machine-readable format, with direct transfer where technically feasible and without harming others' rights (Migrations, Purges & Shutdowns).
• Data residency: processing a region's users' data in that region, with the controller per region, a placement property of accounts and content recorded alongside community placement (Platform Architecture).
• Data Transfer Initiative: a nonprofit since 2023, grown from the 2018 Data Transfer Project, that builds open-source tools and schemas for moving data between services (Migrations, Purges & Shutdowns).
• DAUq (daily active uniques): Reddit's audience metric, a unique identifier that visited a page or opened an app at least once in 24 hours, including logged-out traffic; WAUq is the seven-day version (Operating Metrics).
• Declared automation: an app, bot or AI agent account labeled as such everywhere it speaks, with rate limits and permissions fitted to automation; undeclared automation is inauthentic behavior (AI on User-Generated Platforms).
• Dedicated tier: the placement for communities that outgrow shared resources, the "Big Freaking Guild" move made as a data change triggered by thresholds on members, messages or index size (Platform Architecture).
• Deepfake: in the AI Act, image, audio or video content that a deployer must disclose as artificially generated or manipulated, with a lighter duty for evidently artistic, satirical or fictional work (AI on User-Generated Platforms).
• Deletion order: the published rule for which content goes first under a cap or purge (oldest first, least viewed first, or the user's choice); least viewed minimizes broken links (Migrations, Purges & Shutdowns).
• Deployer and provider (AI Act): the two roles a platform can hold under Article 50: deployer of AI that generates content (disclose deepfakes) and provider of a generative system (mark outputs in a machine-readable format) (AI on User-Generated Platforms).
• Derived store: a search index, feed cache or timeline built from events, the way a platform partitioned by community answers queries that cross communities without scatter-gathering (Platform Architecture).
• Edit and delete race: a concurrent edit and delete under last-write-wins that leaves rows without an author, which every store that merges writes needs a rule for (Platform Architecture).
• Event spine: the event log every derived system reads (search, feeds, notifications, payouts, metrics, federation), such as Reddit's Kafka with more than 500 brokers (Platform Architecture).
• Evidence snapshot: the moderation service's own copy of content as it was judged, kept after edits and deletions under stricter access and retention than the content store (Platform Architecture).
• Exposure metric: a measure of how much violating content users actually saw, such as prevalence, violative view rate or reach before removal, the only kind of trust and safety number that says whether users are safer (Operating Metrics).
• Fan-out: the multiplication of one write into many deliveries, such as a message to a large server or a live chat line to a big audience, the main cost of real-time features (Platform Architecture).
• Fill rate: booked nights over available nights by market and date, which shows where demand exceeds supply (Operating Metrics).
• Frozen period: a published or payout-deciding metric value that is fixed once released, corrected only by a restatement row that supersedes it (Operating Metrics).
• GMV and GMS: gross merchandise value (eBay's term) and gross merchandise sales (Etsy's), the goods marketplaces' volume measures from which take rate is computed (Operating Metrics).
• Gross booking value (GBV): the value of bookings a marketplace counts at booking, adjusted for cancellations and alterations; Airbnb reported $91.3 billion for 2025 (Operating Metrics).
• Hot partition: a partition carrying far more load than others because one community, channel or stream is suddenly busy, the defining skew problem of user-generated workloads (Platform Architecture).
• Inactive account policy: deleting accounts unused for a period after notice, as Google did for personal accounts unused for two years, with several notices to two addresses over months (Migrations, Purges & Shutdowns).
• Link rot: references across the web that break when content is removed, such as Imgur embeds in old Reddit threads; the platform can't prevent it but controls what the URL says afterwards (Migrations, Purges & Shutdowns).
• Liquidity: the chance that a search turns into a transaction, measured as search to booking, acceptance, fill rate and time to first booking; not published by the platforms but what operators run on (Operating Metrics).
• Logged-out user: a visitor without an account, counted in Reddit's DAUq, which needs a stable anonymous identifier and a written deduplication method because counting such traffic is imperfect (Operating Metrics).
• Machine-readable marking: the AI Act duty (Article 50(2)) on generative system providers to mark outputs so they're detectable as artificially generated; systems already on the market before August 2, 2026 have until December 2, 2026 (AI on User-Generated Platforms).
• MAU (monthly active users): the audience unit of a discovery product such as Pinterest, which counts authenticated users active in a 30-day window (Operating Metrics).
• Media provenance: a row read from the original upload at ingest (digital source type, C2PA validation result, signer) that labels every variant at serving time whatever the image pipeline stripped (AI on User-Generated Platforms).
• Metric definition: one versioned definition per metric (unit, population, exclusions such as bots, clock, formula and history policy) that every report reads, so a published figure can be reproduced from its version years later (Operating Metrics).
• Modular monolith: the playbook's starting shape, split into services only when a part needs its own scaling or team, with one owner for every piece of data (Platform Architecture).
• Nights and seats booked: Airbnb's volume unit (533 million in 2025), counted at booking and adjusted for cancellations and alterations (Operating Metrics).
• Partition key: the field data is split by; on a user-generated platform the community (server, subreddit, channel), because reads, permissions and moderation concentrate there (Platform Architecture).
• Payout-linked metric: a number that decides money or status (watch hours toward partner thresholds, subscriber counts, a host's response rate), given versioned definitions, stamped outputs, frozen periods and adjustments (Operating Metrics).
• Policy-driven purge: a bulk moderation or retention decision applied to a population defined by a rule, for policy, security or cost reasons, run in announced, throttled, reversible stages (Migrations, Purges & Shutdowns).
• Pre-send filter: a moderation check on real-time surfaces (chat, live streams) with a tight time budget, such as the 200 milliseconds Amazon IVS gives its chat review handler before a fallback applies (Platform Architecture).
• Prevalence sample: one sampled view, labeled with the policy version and carrying its stratum weight, from which prevalence intervals and classifier precision and recall are computed (Operating Metrics).
• Process metric: a measure of how well moderation works (proactive rate, time to action, appeal overturn rate, report actionability), as distinct from volume and exposure (Operating Metrics).
• Profile move (Mastodon): redirecting followers to a new account after the user sets an alias on it; posts don't move, follows and blocks move by CSV, and another move waits 30 days (Migrations, Purges & Shutdowns).
• Purge policy: a dated bulk removal rule with its versioned population query, reason, optional intermediate state and approver, whose batches are recorded as purge runs that can pause and resume (Migrations, Purges & Shutdowns).
• Purge population: the set of items a purge targets, defined as a versioned query and counted (by community and referrer) before announcement, because the count shows how much of the web will break (Migrations, Purges & Shutdowns).
• Removal propagation: delivering content.removed to every derived store (search, feeds, notifications, AI answer indexes) on a latency target, a moderation feature rather than plumbing (Platform Architecture).
• Response and reach: the share of new posts that get a reply, vote or view within a day, the cold-start metric for creators (Operating Metrics).
• Restatement: a correction to a published or payout number, recorded as a new row that supersedes the old one and published as such, never a quiet change (Operating Metrics).
• Service sunset: ending a feature, service, API version or country as a dated state machine (announced, frozen, read-only, closed, deleted, tombstoned) that users, support and integrations can read (Migrations, Purges & Shutdowns).
• Serving by jurisdiction: carrying the viewer's jurisdiction on every request (network edge, refined by account settings and verified age) and evaluating age rules, geographic restrictions and consent from cached, versioned rule data (Platform Architecture).
• Skew: the concentration of activity in a few communities, posts or streams, which moves as content goes viral, so a quiet partition can become the hottest in minutes (Platform Architecture).
• Tap compare: sending a share of live traffic to a new service and comparing its responses with the old one's while users still get the old responses, as Reddit did moving comments to Go (Platform Architecture).
• Time bucket: a time window in a partition key, such as Discord's roughly ten-day buckets per channel, that spreads one channel's history so no partition grows without bound (Platform Architecture).
• Tombstone (database): a delete marker in stores like Cassandra; Discord found millions of them in one channel caused ten-second garbage collection pauses, which is why deletes and purges are load (Platform Architecture).
• Unit cost: cost per viewer-hour, stored file, delivered image, moderated item or active user, allocated by community and country and tied to the product decision that drives it (Operating Metrics).
• Vendor adapter: the platform-owned boundary around a vendor, designed with a latency budget, a failure mode (fail closed for child-safety scanning) and a record of the data it received (Platform Architecture).
• Violative View Rate: Snap's exposure metric, the share of Snap and Story views that were of violating content, about 0.01% in the second half of 2025 (Operating Metrics).

References#

Reference

• 47 U.S.C. § 230, the immunity and Good Samaritan provisions behind the hosting-liability terms
• 17 U.S.C. § 512, the DMCA safe harbors, designated agent, notices, counter-notices and repeat-infringer terms
• 18 U.S.C. § 2258A, the provider reporting duty behind the CyberTipline terms
• Regulation (EU) 2022/2065 (Digital Services Act), the notice, statement of reasons, trusted flagger and very large online platform terms