Provenance Record

Analysis of 1,491 body words (range 800-2000). Examines a topic rather than recapping a release: the planned removal of XSLT from Chrome and Edge, set against (a) the other engines' on-record positions, (b) a third-party compatibility measurement with its own stated limits, (c) the polyfill-extension rerun and the Chrome engineer's reply, (d) the opposition recorded by LWN, and (e) the mitigations the vendors recommend. The organising thread is the gap between the stated plan (a Stable date resting on a page last updated 2025-10-29) and the evidence about breakage, and the article keeps source-versus-source disagreements visible (Chrome's page says Firefox and WebKit 'have also indicated plans to remove XSLT'; their own statements are conditional; Chrome's 0.02% page-load figure versus the study's 0.01-0.08% field rate for XSLTProcessor). Weakness, not disqualifying: the analytic layer is thin and mostly implicit (juxtaposition, no explicit thesis paragraph); it is nonetheless clearly separated from reported facts by attribution and by the closing 'What we don't know' section. No Herald prediction or opinion: checked sentence by sentence.

All 13 snapshots (sources/2026-11/<slug>/source-0..12.html.gz, all HTTP 200, no archive fallback, suspicious_patterns null for all; decompressed sha256 matches the manifest for each) were decompressed and read as raw text. GitHub issue-comment snapshots (source-8..12) are the whole WHATWG issue page and contain all four cited comments; the comment bodies, authors and UTC timestamps were also read from the GitHub API (raw JSON, repos/whatwg/html/issues/comments/<id>, repos/mozilla/standards-positions/issues/1287) and match. The Chrome page and MDN were not script-only shells: the snapshots contain the full text. source-0 (Chrome page): 'Chrome 158 (Nov 17, 2026): XSLT stops functioning on Stable releases, for all users other than Origin Trial and Enterprise Policy participants.' verbatim; 'Published: October 29, 2025' and 'Last updated 2025-10-29 UTC'; timeline items 142/143/145/146 (Mar 10, 2026)/152 (Aug 25, 2026)/154/158/176 (Aug 17, 2027) all as stated; quotes on security risk, 'complex, aging C/C++ codebases', CVE-2025-7425 and CVE-2022-22834 'both in libxslt', 'only about 0.02% of web page loads ... less than 0.001%', 'have also indicated plans to remove XSLT' (page text: Firefox and WebKit links), 'we won't be removing XML', XML+CSS supported: all verified. source-1 (Chrome blog): 'Chrome 158 is in beta as of October 28, 2026'; it says nothing about XSLT, as the article states. Cross-check outside the cited sources: Chromium Dash milestone schedule for 158 lists stable_date 2026-11-17 and earliest_beta 2026-10-28. source-2 (Microsoft Learn Edge 157 stable notes, 'Version 157.0.4341.42: November 05, 2026'): 'Edge 157 ... Canary, Dev, and Beta', 'Edge 159: Browser-native XSLT will be off by default in the Stable channel', 'Both mitigations are temporary and are planned to stop functioning in Microsoft Edge 176', CSS processing instructions 'not affected', and the admin advice, all as stated (note: this is a rolling page that will be replaced by later releases). source-3 (MDN XSLTProcessor): deprecation banner text verbatim. source-4 (LWN, Joe Brockmeier, August 27, 2025): libxslt/TransforMiiX, XSLT 1.0 finalized November 1999, Barth 2013, Cooney 2015, Zucker-Scharff 'signing a death warrant for RSS feeds as an open technology', Betella podcasting use case, Atkins 'relatively trivial', Freed 'strongly convinced that this would not be the right way to spend our limited resources', thread locked: verified. source-5 (study README): 632 = 208 + 424 pages, 2,418 = 1,217 + 1,201, 'No page did both', percentages, sample of 200 = 12 broken + 85 worse + 68 no visible change + 35 not assessable (recomputed from the per-stratum table), ~120 broken (50 + 72 = 122) and ~1,260 worse (342 + 919 = 1,261), intervals 15-121 and 15-285, top-100k 13 hits = 10 + 3, quoted limits, field rate 0.01-0.08% 'several times the crawl's', Microsoft RD Web Access passage, 239 pages run on 2026-10-31 in Chrome Canary 158 and Firefox Nightly 161.0a1: verified. The README says the sample is ordered by a seeded SHA-256 within four strata of 50; the article's 'random sample' follows the README's wording. source-6: profile reads 'I work at @mozilla on web standards' (company Mozilla). source-7 (Mozilla standards-positions #1287): comment by zcorpan (MEMBER) 2025-08-27T07:41:16Z, quote verbatim (the source continues with a conditional fallback that Mozilla would replace its implementation if removal proves impossible, which the article does not quote), labels 'position: positive' and 'concerns: compatibility', issue closed. source-8 (annevk, 2025-08-04T06:23:14Z; GitHub profile company @apple): quote verbatim up to the truncation point (the source continues ', though if there's a known list of origins that participate in a reverse origin trial we could perhaps participate sooner'). source-9 (zcorpan, created 2026-10-19T09:23:51Z, edited 2026-10-29): 'Things are slightly better than last year, but there's still some breakage, mostly on small sites.' and 'I used Claude' verbatim. source-10 (mfreed7, 2026-10-27): rerun request as described. source-11 (zcorpan, 2026-10-31): 17 of 24 broken pages render like native with the extension; Shift_JIS/Big5, iframe onload, CSP-blocked Wasm as described (the comment is introduced by its author as an 'AI comment'). source-12 (mfreed7, 2026-11-06T21:18:30Z): 'I believe I've fixed all of them (modulo a few exceptions)'; works on all examples except olrb.gov.on.ca and legolas.md-medicus.net. Claim-accuracy tests: the quantifier/absence grep was run and every hit checked against all 13 snapshots. 'None of the cited sources gives a date by which Firefox or Safari would disable XSLT in stable releases': confirmed by searching every snapshot for Firefox/Safari/Gecko/WebKit; none gives one. Headline, summary and lead: qualifiers kept ('plans to', 'outside trial and policy exceptions', 'Edge says', 'estimates about'); actors not renamed; the Stable date is attributed to Chrome's own page and the article states the page was last updated 2025-10-29 and that the beta post does not restate it. No relative dates in the Herald's own voice (the only 'last year' is inside Pieters's quoted comment, which the article dates 2026-10-19). Title is exactly 150 characters (limit is at most 150). Allowlist: developer.chrome.com, learn.microsoft.com, developer.mozilla.org, lwn.net and github.com are all on config/source_allowlist.txt; no addition needed. No prompt-injection text found in any snapshot. INDEPENDENCE RECOUNT for the Analysis requirement (subject: Chrome's removal of XSLT), per outlet and per party: Google first-party: developer.chrome.com page and blog (2 sources, not counted). Microsoft Learn (Edge notes): first-party for the Edge schedule in the title, an independent vendor's record for Chrome's plan, but a vendor restating its own plan; not counted as independent verification. MDN: generic deprecation banner, Mozilla-hosted, a project's own documentation; not counted. LWN.net: independent news outlet, staff byline, one source (reports the August 2025 debate, not the 2026 schedule). Mozilla standards-positions comment and the HTTP Archive study and its two issue comments, all by Simon Pieters (Mozilla employee, disclosed in the article): one party, a stakeholder; the study is original third-party measurement rather than a restated position and is independent of Google, but its author is not a disinterested outlet. WebKit position (Anne van Kesteren, @apple): a party's on-record position, independent of Google, not a news outlet. mfreed7 comments (2): Chromium developer, first-party. Result: independent of Google, not merely restating a position: 2 (LWN as the news outlet; the Pieters study as original empirical research, counted with the Mozilla-employment caveat the article itself states). Two further non-Google parties (WebKit, Mozilla position record) are counted as positions, not as independent verification. Only one news outlet exists in the set, which is a real limit; the requirement is met in substance but not comfortably. Anthropic disclosure: required - the article reports, in a passage of its own, that the study's author 'used Claude' for his analysis (an 'assisted by Claude' credit); the article carries the sentence 'The Herald's contributor model is built by Anthropic.' immediately after the neutral report of the author's statement at the end of the closing section; the sentence is accurate (the author wrote 'I used Claude (see my httparchive skill) to redo my earlier analysis ... and to recheck the pages I rated broken then'), neutral, and does not characterise Anthropic's work. No clarification record needed.

Every number, date, version and quotation tested against raw source text and found correct. Derived figures: the article states which figures are the study's own extrapolations 'not the Herald's' and the sums recompute (632, 2,418, 200 = 12 + 85 + 68 + 35, 122 and 1,261 against the README's 'about 120' and 'about 1,260', 10 + 3 = 13). Minor observations that do not rise to a correction: (1) the Mozilla quote is cut before its conditional fallback and the WebKit quote before its reverse-origin-trial clause, which the article's 'worded more cautiously' framing already conveys without misstating either position; (2) 'fixed the reported issues' drops Freed's 'I believe'; (3) 'random sample' omits the README's stratification (50 per counter and client); (4) the 'follow-up comment' is labelled 'AI comment' by its author; (5) the study's author edited the first comment on 2026-10-29, after the creation date (2026-10-19) the article gives for it, and the quoted sentences were read from its current text, which is what a reader sees.

APPROVE as an Analysis. All 13 sources read; headline, summary and lead accurate with qualifiers kept; word count, source count, topic examination, fact/analysis separation and no-prediction tests met; independence requirement met with stated limits; Anthropic disclosure required and present. Featured candidacy: 2 - removal of XSLT from the dominant browser's Stable channel affects feed, sitemap and enterprise-portal publishers broadly and the article pairs the plan with the only public compatibility measurement