An RSS feed is a machine-readable file, written in XML, that lists the most recent items a website has published. Each entry carries a title, a link, a date and usually a summary or the full text. Software fetches the file on a schedule, compares it with the previous copy and surfaces what is new. The format exists so that following a publication does not require visiting it. The subscriber's software does the checking, the publisher decides what appears in the list, and no intermediary sits between the two deciding which items are worth showing.

The initials are unstable. The dominant reading is Really Simple Syndication, the name Dave Winer attached to the 2002 revision. Parallel lineages used RDF Site Summary and Rich Site Summary for formats sharing a common ancestor but diverging in structure. All three still appear in documentation.

Inside the file

At the top sits an rss element with a version attribute. Below it, exactly one channel element describes the publication. The RSS Advisory Board specification requires three child elements on that channel: title, link and description. Everything else is optional or arrives by namespace.

Each entry is an item element, in practice carrying title, link, description, pubDate and guid. The date follows the RFC 822 convention: day of week, day, month, year and a 24-hour time with a time zone offset. The guid is a globally unique identifier that must never change, a rule Apple's podcast requirements repeat, because a rewritten identifier makes an old episode look new and duplicates it in every subscriber's app.

The element that turned a text format into an audio industry is enclosure. It is empty, a single tag carrying three required attributes: url pointing at the media file, length giving its size in bytes, and type carrying the MIME media type, typically audio/mpeg. UserLand added it in RSS 0.92 on 25 December 2000. The specification advises against more than one enclosure per item, because aggregator support for multiples is inconsistent, and Apple Podcasts ignores duplicate enclosure URLs outright.

Namespaces carry everything the core format omits. RSS 2.0 elements deliberately sit outside any namespace, which is what keeps a 0.91 file valid as a 2.0 file, while extensions arrive prefixed. The iTunes namespace, declared as http://www.itunes.com/dtds/podcast-1.0.dtd and introduced alongside Apple's podcast directory in 2005, supplies itunes:category, itunes:explicit, itunes:author and itunes:image. Feeds missing those tags fail Apple validation, and artwork must sit between 1400 and 3000 pixels square at 72 dpi. Media RSS, designed by Yahoo in 2004 and handed to the RSS Advisory Board in 2009, extends enclosures with thumbnails and multiple renditions of one asset, which is how video libraries reach players and free ad-supported streaming channels.

Discovery runs through a link element in the HTML head, with rel set to alternate and type set to application/rss+xml, a media type never formally registered. A reader handed a homepage address looks for that tag instead of guessing at file paths. Transport is polling: nothing is pushed. Readers request the file at intervals and rely on conditional requests, sending If-Modified-Since or If-None-Match headers so an unchanged feed returns a 304 rather than the full document.

From Netscape to a frozen specification

Web syndication predates the acronym: Channel Definition Format, Meta Content Framework and Winer's scriptingNews all described site contents in XML during 1997.

Netscape published RSS 0.90 on 15 March 1999, authored by Ramanathan Guha, to populate its My Netscape portal. Dan Libby's RSS 0.91 followed on 10 July 1999, dropping the RDF layer. Netscape then lost interest. UserLand issued its own 0.91 on 9 June 2000, added enclosures that December, and released RSS 2.0 on 19 August 2002. A separate group published the RDF-based RSS 1.0 in December 2000, and the fork has never been resolved.

Copyright in the 2.0 specification passed to Harvard Law School's Berkman Klein Center, which published version 2.0.1 on 15 July 2003. The RSS Advisory Board took over maintenance the same week, and its last revision, version 2.0.11, appeared on 30 March 2009. Nothing has changed since. The Internet Engineering Task Force standardised the competing Atom format as RFC 4287 in December 2005, partly to fix the ambiguities the frozen RSS text left behind.

Where advertising money touches the feed

Podcast advertising is built on the enclosure element, and the commercial layer sits in the URL inside it. Measurement firms operate prefix redirects: the publisher prepends a third-party domain to the media URL, so every download passes through that party's server before reaching the file. Server logs, not client-side beacons, are the record of delivery. Dynamic ad insertion works at the same choke point, stitching a different advertisement into the file on each request.

That mechanism defines what can be counted. IAB Tech Lab released version 2.3 of its Podcast Technical Measurement Guidelines on 21 July 2026, a 32-page draft that extends download and ad-delivery rules to video episodes travelling through open feeds, documents the prefix redirect method in section 5.2, and replaces the word listener with podcast consumer. Comments closed on 19 August 2026. Version 3.0, aimed at streaming video podcasts, is scheduled for 2027. The draft also defines three separate measurement window types, which can produce three different totals from identical server logs.

The byte thresholds are asymmetric. An embedded advertisement in the first quarter of an episode counts as delivered once 25% of the file has been served, while a dynamically inserted spot requires the full advertisement to transfer. A podcast download is a different metric from a display impression counted on download, which the Media Rating Council ruled against for display advertising two decades ago.

The open feed boundary is now a commercial dividing line. Episodes streamed inside Netflix do not travel through an open RSS feed and are not delivered by a hosting company's server, placing them outside the version 2.3 framework. Podtrac found that podcast playlist episodes on YouTube retain 64% of counted views as engaged views, a discount that open-feed download rules do not address.

Why it matters beyond podcasting

Feeds remain the plumbing under several systems marketers use without naming them. Google's Feedfetcher crawls RSS and Atom feeds for Google News and WebSub under its own user agent string. Enterprise podcast platforms such as Omny Studio treat feed management as a core hosting function alongside ingestion and analytics. YouTube offers creators a feed integration path in which one update propagates to every directory at once.

Competitive monitoring is the other durable use. Inoreader, founded in Sofia in 2013 after speculation that Google Reader would close, added podcast transcripts in September 2025. Openvibe added feed support in August 2025 to pull blogs and newsletters into the same timeline as Mastodon, Bluesky, Nostr and Threads. Evan Schwartz built Scour on semantic matching across 12,903 sources, exporting every personalised feed as RSS, Atom and JSON.

Limitations and contested ground

A feed carries no engagement data. There is no identifier, no cookie, no session and no way to know whether an item was read. Polling wastes requests on unchanged files, and full-text feeds hand content to anyone who fetches them, which is why publishers often truncate.

The sharper dispute is infrastructural. Google Reader closed in 2013. RSS support disappeared from Google Groups in 2021, and Google News stopped using feeds submitted through Publisher Center in March 2025, after the rebuilt Publisher Center had already moved publishers towards section URLs. Andrew Dolgov shut down the self-hosted reader Tiny Tiny RSS in November 2025.

Then came XSLT. On 20 August 2025, Google engineer Mason Freed proposed removing the transformation language from browsers through WHATWG, the mechanism that renders raw feed XML as a readable page. Developers documented breakage across the United States Congress, the National Weather Service, the European Parliament and Australian and Canadian agencies. Google's rationale cites memory-safety vulnerabilities in libxslt and usage below 0.02% of page loads, with the processing instruction under 0.001%. Mozilla and WebKit signalled support. Critics read a pattern rather than a security fix.

Not the same as

Atom is a competing syndication format, standardised by the IETF rather than a private board, with stricter rules on identifiers and dates. Most readers accept both.

A product feed is not an RSS feed. In search and shopping advertising, feed usually means a structured inventory file supplied to Google Merchant Center or a comparable catalogue, which shares nothing with syndication beyond the word.

A sitemap is also XML, but addresses crawlers and lists every URL on a site rather than the newest items in reverse chronological order. WebSub is the push counterpart to polling, notifying subscribers when a feed changes instead of waiting for the next fetch.

Recent developments

Chrome's origin trial for continued XSLT support opened on 25 August 2026. Removal from stable releases follows on 17 November 2026, with the origin trial and enterprise policy escape hatches closing on 17 August 2027. Google's documentation names those releases as Chrome 158 and Chrome 176; the earlier intent-to-remove filing on the Blink developer list gave the same dates against milestones 155 and 164.

Podcast distribution keeps testing the boundary. Spotify activated its video distribution API across five hosting platforms on 14 May 2026, stating that the feed remains the main channel to audio-only services and that a show can be moved back to audio-only distribution. Apple's HTTP Live Streaming route preserves the open feed while adding impression verification that plain file delivery never offered. Of 34,208 video podcasts surveyed by MillionPodcasts, 77% carry no advertising at all.

Timeline

  • 1997: Channel Definition Format, Meta Content Framework and scriptingNews propose XML syndication
  • 15 March 1999: Netscape publishes RSS 0.90, authored by Ramanathan Guha
  • 10 July 1999: Netscape publishes RSS 0.91, authored by Dan Libby
  • 16 November 1999: W3C publishes XSLT 1.0 as a recommendation
  • 9 June 2000: UserLand publishes its own RSS 0.91
  • December 2000: RDF-based RSS 1.0 published by a separate working group
  • 25 December 2000: RSS 0.92 adds the enclosure element
  • 19 August 2002: UserLand publishes RSS 2.0
  • 15 July 2003: Berkman Klein Center publishes version 2.0.1; the RSS Advisory Board forms
  • 2004: Yahoo designs Media RSS
  • 2005: Apple introduces the iTunes namespace with its podcast directory
  • December 2005: IETF standardises Atom as RFC 4287
  • 30 March 2009: RSS 2.0 version 2.0.11 published, the last revision to date
  • 2009: Media RSS maintenance transfers to the RSS Advisory Board
  • 2013: Google Reader closes
  • December 2018: Firefox removes built-in feed reading and live bookmarks
  • 2021: Google removes RSS support from Google Groups
  • March 2025: Google News stops using feeds submitted through Publisher Center
  • 20 August 2025: XSLT removal proposed through WHATWG
  • 29 October 2025: Chrome publishes its XSLT deprecation timeline
  • 1 November 2025: Tiny Tiny RSS infrastructure shut down
  • 21 July 2026: IAB Tech Lab opens version 2.3 of its podcast measurement guidelines
  • 25 August 2026: Chrome origin trial for XSLT opens
  • 17 November 2026: XSLT stops functioning on Chrome stable releases
  • 17 August 2027: Chrome origin trial and enterprise policy expire

Summary

Who: Created at Netscape by Ramanathan Guha and Dan Libby, developed further by Dave Winer at UserLand, and maintained since 2003 by the RSS Advisory Board under a copyright assigned to Harvard Law School's Berkman Klein Center. Apple, the IAB Tech Lab, podcast hosting companies and independent reader developers extend and depend on it.

What: An XML document listing a publication's most recent items, each with a title, link, date, identifier and optional media enclosure, fetched on a polling schedule by subscriber software.

When: First published on 15 March 1999, extended with enclosures on 25 December 2000, frozen at version 2.0.11 on 30 March 2009, and subject to browser rendering changes taking effect on 17 November 2026.

Where: Across podcast distribution, news aggregation, government publishing, and the feed-driven ingestion paths of platforms including Google News, YouTube and Spotify.

Why: Podcast advertising is measured and monetised at the enclosure URL, so the format determines what can be counted and sold in a channel with no cookies or device identifiers. For publishers, a feed is the one distribution route no platform ranks, throttles or personalises.