W3C's WCAG 3 resolution did not freeze the text

Summary

On 11 August 2026, W3C's Accessibility Guidelines Working Group announced “consensus with an objection” on keeping WCAG 3's normative conformance section focused on conformance while moving advice for regulators mainly into informative guidance. People who had missed the meeting received five working days to add objections.
The group continued debating the scope of conformance on 18 August. The chair said no resolution would be reached that day, and the meeting ended that part of the discussion with an exploratory straw poll rather than frozen language.
On 25 August, AGWG adopted three resolutions to update sections 3, 3.1 and 3.2 of the Editor's Draft using text in a linked Google document after resolving associated comments. The conditional words matter: the decision authorized direction and further bounded editing.
Official pull request 826 remained active after the meeting, accumulated post-meeting integration, editorial and feedback commits, and merged on 26 August. The public Editor's Draft dated 28 August still says it is work in progress, may be replaced and is not W3C or Member endorsement.
A group resolution, a merged Editor's Draft, a W3C Decision and a final Recommendation are different authority states. A compact decision-to-diff concordance should keep them joined without pretending that a repository created the mandate or that the meeting approved words not yet settled.

The decisive phrase came after the link

The 25 August minutes contain an unusually candid formulation. Three times, the Working Group resolved to update a part of the WCAG 3 conformance material with text in a linked Google document—after resolving associated comments in Google Doc.

The resolutions covered the developing conformance section, the subsection on interpreting normative provisions and the subsection on conformance requirements. They did not merely say that editors should think about the subject. They authorized concrete changes to the public Editor's Draft. Yet they also did not identify one immutable document digest and declare every character final. The text still contained comments that the authorized editing process had to resolve.

That distinction is not a flaw hidden between the lines. It is written into the act. A conditional drafting resolution can be responsible governance when a long text has already been surveyed and discussed but still needs precise comment disposition. The error would come later, if a reader treated the merged document as though every word had been displayed and approved during the meeting.

One participant raised almost exactly that record problem. The minutes quote a question about whether the group could decide language in a Google document rather than language reproduced in the IRC record. The response was practical: the material was too long to fit comfortably in IRC. That is a reasonable constraint. It does not eliminate the need for a durable identity for the reviewed text.

The decision had a two-week prehistory

The 25 August act was not made from a standing start. On 11 August, AGWG announced that it had reached consensus with an objection on a basic allocation of subject matter. The normative conformance section would focus on conformance, including a scoped conformance claim. Recommendations to regulators about compliance for content not fully within an author's control would sit in informative guidance and possibly in particular WCAG provisions.

The objection was summarized publicly: this was a large decision while the group had not yet defined conformance. The announcement invited people who had not attended to add objections within five working days. It then promised one of two next states: record the decision, or bring it back to the group if needed.

The language must be read carefully. “Consensus with an objection” is not unanimity. It also is not a statement that one objection automatically blocks all work. W3C Process allows a chair to record a group decision where dissent remains after the group has considered legitimate concerns as far as reasonably possible. A dissenter may elevate a sustained objection through the Formal Objection process. Those are distinct events, with different evidence and different consequences.

Nor did 11 August settle the text. On 18 August the group devoted another substantial discussion to conformance scope. The minutes preserve competing views about pages, paths, processes, components, claims and reporting. Proposals were drafted and redrafted during the call. Near the end, the chair stated that the group would not get a resolution that day. A straw poll supported continuing to explore a more limited conformance section with defined scopes and informative documents. Exploration was the state; final wording was not.

This chronology changes the meaning of the later act. The 25 August resolutions were the next authorized transition after a direction, an objection window, renewed deliberation, a survey and draft text. They were not a sudden plebiscite and should not be described as one.

A resolution is not a vote merely because the minutes contain plus signs

The 25 August record shows participants expressing +1, 0 and -1. That notation is useful evidence of the room. It does not automatically make the procedure a formal vote.

W3C Process distinguishes consensus-building from voting. A substantive vote is a last-resort device after the chair determines that technical discussion and compromise have failed; the decision to vote, the voting rule, the outcome and any Formal Objections must be recorded. The minutes at issue instead use the language of draft resolutions, participant signals and committed resolutions. This article therefore calls them group resolutions, not ballots.

The distinction matters because counting visible signals can produce an attractive but false model of authority. Participants provide arguments, objections, experience and support. The chair assesses whether a group decision exists under the process. The group can authorize work within its charter and process. None of that converts attendance into authority over regulators, courts, public procurers or every person who uses the web.

The W3C Process Document supplies another necessary boundary. It separately defines chair decisions, group decisions or resolutions, Team Decisions and W3C Decisions. A Working Group resolution is an institutional act, but it is not the final W3C-wide act created after the relevant Recommendation-track procedures. Giving the resolution its proper weight protects it from two opposite distortions: dismissing it as mere conversation, or inflating it into a completed standard.

The repository shows execution after the decision

The clearest evidence of subsequent execution is official pull request 826 in the w3c/wcag3 repository. It was opened before the 25 August meeting under the title “Update conformance section.” It merged the next day, on 26 August at 23:57 UTC.

Between opening and merge, the public history shows a long sequence rather than one frozen upload. Commit descriptions include an initial conformance update, removal of section 4, editorial adjustments, integration of post-meeting conformance content, refactoring, changes to the plain-language summary and introductory editor's note, targeted review edits and a final address feedback commit. The final branch head and the merge commit have stable identities.

Those commits do not prove deviation. The resolutions expressly anticipated comment resolution. Several commits may be purely editorial or mechanical; others may implement points already considered in the survey and meeting. A commit message is not a constitutional judgment. The public record reviewed for this article does not establish that any editor exceeded the delegated task.

What the history proves is narrower and more useful: the exact merged text was produced through implementation after the group decision. The repository is excellent evidence of that execution. It is not the source of the group's authority. Git can show which bytes changed and who committed them. It cannot, by itself, answer which changes the resolution permitted, which comment each change closed, or whether later group review superseded an earlier direction.

The database is not a throne, and neither is a merge button.

The Editor's Draft tells readers not to treat it as final

The public Editor's Draft dated 28 August makes its own status unusually difficult to misunderstand. Publication as an Editor's Draft does not imply endorsement by W3C or its Members. The document can be updated, replaced or made obsolete at any time. It is inappropriate to cite it as anything other than work in progress.

The conformance section is marked Developing. In the draft's status vocabulary, that means the group has general agreement on the topic while details remain unfinished. That is compatible with a real 25 August group resolution. It is incompatible with describing the text as a final W3C Recommendation or a rule already imposed on the outside world.

The latest formally published WCAG 3 Working Draft and the newer Editor's Draft serve different documentary purposes. The Editor's Draft exposes current work. A formal publication records a later reviewed snapshot. A future Recommendation would carry another status again. Regulators and other external institutions may incorporate a W3C standard through their own legal or contractual authority, but the Working Group cannot make that external adoption happen by discussing regulatory guidance.

What a decision-to-diff concordance would add

No one needs a transcript attached to every comma. A useful record can remain compact.

Decision identity

It should give the issue and resolution stable identifiers and name the decision type. It should preserve the exact resolution, the meeting or asynchronous window, the eligible decision set and the time the chair declared the result.

Reviewed text

It should identify the text actually reviewed. For a repository-native draft, that may be a commit. For a collaborative document, it may be a versioned export and content hash. The record should name the sections and anchors covered, not merely the title of a large document.

Open comments

It should enumerate the comments or issues left for post-decision resolution. The public does not need personal annotations or protected member material. It does need a bounded statement of what remained open and who was authorized to resolve it.

Change classification

The concordance should classify the post-decision change set. A commit can implement the resolution, resolve a named comment, make an editorial correction, repair a build problem or change scope. The label does not decide whether the change was wise. It lets a reviewer find the relevant authority and evidence.

Final state

The record should name the final head and merge commit, publication state and any later Call for Consensus, objection disposition, reopening or superseding decision. A machine-readable diff from the reviewed snapshot to the merge would make the join reproducible.

Authority boundary

It should state the authority boundary. The resolution authorized a Working Group drafting transition. The merge executed it. The Editor's Draft displayed a current result. None of those states alone was a W3C Recommendation, a government rule or proof of accessibility in a particular product.

The public index is a projection, not the act

AGWG maintains a public Decisions page. When checked for this article, it showed a last-edit time of 17 August and did not yet list the 25 August section-3 resolutions. That observation must not be overread. An index that predates an act cannot prove that the later act was invalid, concealed or forgotten.

It does, however, illustrate why state projection matters. The meeting minutes, decision index, pull request, Editor's Draft and formal publication page each answer a different question. Readers should not have to infer the authoritative join from dates and link trails. Once the index is updated, it can point to the exact resolution and merge. Until then, the minutes remain the public group record and the repository remains the public implementation history.

The right response is reconciliation, not accusation.

W3C Resolved to Change WCAG 3. It Did Not Freeze the Text

Why it matters