Summary

  • The W3C Math Working Group's 2026 charter makes every face-to-face or teleconference resolution provisional, including publication decisions, until a call for consensus has run.
  • The response period is one week, subject to the Chairs' evaluation. If no objection is raised before it closes, the charter says the resolution is considered Working Group consensus.
  • The operative W3C Process supplies the other half of the test: consensus needs substantial active support and no sustained objection. Silence is an abstention, not an affirmative voice.
  • Public mailing lists, GitHub issues and public minutes widen scrutiny, but the default set eligible to participate in a group decision is the set of Working Group participants.
  • A compact consensus receipt should record what text was proposed, who was eligible and notified, how much active support existed, what remained silent, how objections were disposed of, what the Chairs determined and whether new information reopened the decision.

The meeting does not finish the decision

A resolution can leave a W3C Math Working Group meeting with everyone in the room believing the matter is settled. Under the group's 2026 charter, it is not.

Every resolution adopted at a face-to-face meeting or teleconference remains provisional. Publication decisions are included. A call for consensus, or CfC, must then be issued through a persistent channel such as email, a GitHub issue or a web survey. The response period is one week, depending on the Chairs' evaluation of the group's consensus. If the period closes without an objection, the resolution is considered to have Working Group consensus.

That is a thoughtful institutional design. Meetings favour people who can attend at the selected time, speak quickly and judge a proposal while the discussion is moving. An asynchronous week gives an absent participant time to read the exact proposition, consult colleagues, test a technical consequence or turn a hesitation into a reasoned objection. The charter expressly describes the mechanism as a way to afford asynchronous decisions and organisational deliberation.

The public record also gains a useful boundary. A meeting minute can show what the people present resolved. A later CfC can show whether the larger eligible group allowed that resolution to become a group decision. The two events should not be collapsed into one timestamp.

The weak point is not the week. It is the temptation to let one observable condition—no objection—stand in for the whole meaning of consensus.

No objection and active support are different evidence

The Math charter's closing rule is short: if no objections have been raised by the end of the response period, the provisional resolution is considered to have consensus. Read alone, that sentence can make a quiet list look conclusive.

The charter cannot be read alone. It says the current W3C Process takes precedence if the two conflict, and its decision policy expressly invokes the Process definition of consensus. The 18 August 2025 Process defines consensus as a state in which a substantial number of eligible individuals support a decision and nobody in that set sustains an objection. It then says that silence is an abstention.

These are two independent evidentiary dimensions:

  1. Was there a sustained objection?
  2. Was there substantial active support?

A silent participant supplies neither an objection nor support. The silence can help satisfy the first dimension, because there is no expressed objection from that person. It cannot be counted as affirmative evidence for the second.

The Process permits “lazy consensus”, in which lack of objection after sufficient notice is taken as assent. That procedural permission is real. It lets groups move without forcing every participant to answer every routine call. But the same Process warns against decisions marked by widespread apathy, meaning little support and many abstentions. It advises groups to set minimum thresholds of active support before recording a decision. It does not prescribe one universal percentage.

The result is not a contradiction. Lazy consensus is a method of obtaining and assessing consensus. Silence remains an abstention inside the definition. The Chair still has to judge whether the observable support is substantial enough for the decision and context at hand.

A one-week clock cannot supply the denominator

The public Math charter expects the group to maintain six or more active participants over its duration, including representatives of key implementors and active editors and test leads. That is an operating expectation, not a quorum rule for each decision. The charter adds no numerical quorum and no active-support threshold to its one-week CfC clause. The W3C Process likewise requires no default quorum or fixed percentage, though a charter may add either.

Flexibility is defensible. A correction to an editorial error, a publication transition and a change affecting browser behaviour do not necessarily deserve the same participation threshold. A small specialised group can also be damaged by a rigid rule that lets two unavailable participants freeze routine work.

But flexibility places more weight on the decision record. “No objections received” answers only what arrived at the objection channel. It does not say how many participants were eligible, which of them received notice, how many actively supported the proposal, how many explicitly abstained, or how many did nothing. It does not reveal whether the Chair saw broad engagement or merely an empty inbox.

This is the denominator problem in miniature. Lu Heng's critique of multi-stakeholder authority is larger than a standards working group, but its discipline applies: participation, advice, objection and mandate must not be rhetorically merged. A decision can be perfectly valid inside a delegated technical body without being described as though every silent person endorsed it. The narrower the actual authority claim, the stronger the record becomes.

The Math Working Group does not need a plebiscite. It needs a receipt that makes its own decision rule reproducible.

Public channels do not make the public the decision set

The charter gives the public substantial visibility. Technical discussions are public. Teleconference and face-to-face minutes are archived for public review. Issues can be read and written through public channels. The group primarily works on the public www-math list and GitHub, and non-participants are invited to make technical contributions under the relevant patent-policy conditions.

The meetings themselves are not open to public participation. More importantly, the W3C Process says that, by default, the people eligible to participate in a group decision are the group's participants. A public commenter may supply decisive technical evidence without becoming a voter or a member of the consensus set. A participant may be eligible yet remain silent. Those positions carry different procedural meanings.

A robust receipt should therefore preserve two maps rather than flattening them. The contribution map shows who offered a comment, test, objection or alternative, including people outside the group. The decision map shows the eligible participant set and the Chairs' assessment of support and dissent within it. Public technical merit should be considered; formal decision eligibility should be accurately described.

This separation protects openness from two opposite errors. One error dismisses an external review because the reviewer is not a participant. The Process requires Chairs to consider legitimate views and objections from inside and outside the group. The other error presents an open issue tracker as if every account on it were part of the Working Group's decision mandate. Openness is an evidence surface, not an undifferentiated electorate.

The Chair's judgment is real authority

W3C distinguishes a personal Chair decision from a group decision that the Chair records after assessing group consensus or conducting a vote. The Math CfC belongs to the second category. Its result is not produced automatically by a timer. The Chairs evaluate the group consensus, and the Process says the final determination remains with the Chair when questions or disagreements arise.

That authority is necessary. Objections vary in strength. A participant may disagree but accept the proposal. A sustained objection may expose a technical failure. Several messages may come from people at the same member organisation. A late comment may contain information that was unavailable during the meeting. Counting messages cannot resolve those distinctions.

Judgment, however, should leave an inspectable trail. If the decision closes, the record should say what the Chairs assessed as substantial support and why any dissent did or did not prevent consensus. If a vote becomes necessary, the Process requires the issue, decision to vote, outcome and Formal Objections to be recorded. If new information reopens a decision, the reopening should also be recorded, and must be recorded when a participant requests it.

The record need not publish private correspondence or turn every Chair explanation into a legal brief. It should expose enough state to distinguish a reasoned consensus assessment from the mechanical expiry of a date.

What a consensus receipt would contain

The first field is the proposal identity. A CfC should point to the exact resolution, document version, pull request or publication state under consideration. If the text changes materially during the week, the record should say whether the existing call still applies or a new call begins. A title alone is not a version.

The second field is the eligible set. The receipt can record the number of current group participants, not their private contact data. It should state which recognised channels carried the notice, when the call opened, when it was scheduled to close, and whether the Chairs varied the ordinary one-week period.

The third field is participation state. At minimum, distinguish affirmative support, explicit abstention, silence, disagreement that was not sustained, and sustained objection. Public comments by non-participants belong in the evidence section rather than being discarded or counted as group votes.

The fourth field is disposition. Each substantive objection should link to the response, change, no-change rationale or remaining Formal Objection. The W3C Process already expects substantive issue responses and their rationales to be recorded. The receipt only joins that material to the decision whose legitimacy depends on it.

The fifth field is the decision. It should identify the deciding Chairs, the close time, the assessed consensus state and the resulting resolution. If the Chairs relied on a threshold or contextual test, state it. If no numerical threshold applied, say that rather than inventing one after the fact.

The final field is lineage. A later reopening should point back to the original decision and identify the new information, new discussion and successor decision. The earlier record remains historically true; it should not be overwritten as though the first call never occurred.

This receipt would not create a popular mandate, bind people outside W3C or replace the Chairs. It would make the transition the Chairs already control visible enough for another reader to reconstruct.

A comparison shows the record is feasible

The WCAG Working Group has published a more explicit consensus procedure. Its page says a teleconference resolution may precede a CfC but cannot replace it; a call points to relevant discussions; objections can be recorded, addressed or cause discussion to reopen; and formal decisions are placed on a decision page.

That procedure is not a rule for W3C Math. The page is old, applies to a different group and expressly warns that it may not represent W3C consensus. Its value here is narrower: it demonstrates that linking a proposal, supporting discussion, objections, disposition and final decision is operationally possible within W3C's working culture.

Math should choose a record proportionate to its own workload. A structured GitHub issue, archived email summary or generated decision index could be enough. The design target is not more ceremony. It is less ambiguity at the exact point where provisional becomes resolved.

Evidence boundary

No checked source shows that a particular Math CfC lacked active support, omitted a participant, mishandled an objection or closed improperly. This Article audits the published rule, not an undisclosed set of outcomes. It does not allege a breach of the W3C Process.

Nor does the Process turn silence into a veto. Participants may abstain, and groups must be able to advance work. Consensus is not unanimity. A group can record a decision in the presence of dissent after considering legitimate concerns, and it can vote when technical discussion and compromise cannot break a deadlock.

The provable boundary is simpler. The Math charter names no objection as the close condition for a one-week CfC. The governing Process defines silence as abstention and consensus as substantial support plus no sustained objection. A responsible public record should preserve the join. The clock can tell a group when to decide. It cannot speak for anyone who did not.

Sources

  1. Lu Heng, “The Policy Mirror”
  2. Lu Heng, “The Multi-Stakeholder Mirage”
  3. W3C, Math Working Group Charter (2026)
  4. W3C Process Document, 18 August 2025
  5. W3C, Math Working Group public group page
  6. W3C Math home page
  7. W3C, Working/Interest Group Chair guide
  8. W3C, “W3C updates its Process Document”
  9. WCAG Working Group Consensus Procedure