Summary

  • draft-eggert-appeal-support-01, published on 20 July 2026, would require an appellant to obtain at least three qualified supporters before an appeal is considered by the IESG as a whole or the IAB. It does not impose that gate on the initial dispute with Working Group chairs or the responsible Area Director.
  • The draft defines support narrowly: a supporter agrees that the review body should spend time on the matter. The supporter does not thereby agree or disagree with the appellant's position. Three is an admission threshold, not a vote, a finding, rough consensus or institutional endorsement.
  • Supporters must be NomCom-eligible under RFC 9389 as of the appeal date, must not have supported the same appellant during the preceding year, and generally must file personally within two weeks. A qualified appellant may self-support once; an escalation from the IESG to the IAB needs a fresh set.
  • The handling body may decline an insufficiently supported appeal, but may also hear it anyway. If this proposal advances, a privacy-bounded admission receipt should make the count, qualification rule, time window, discretion and escalation reset reviewable while keeping the later merits decision separate.

Three tokens on a closed file

The cover sheet is where institutional language starts to drift. A clerk sees three names and writes “supported appeal.” A dashboard shortens that to “support: 3.” A later history says the appeal had community support. By the time the matter reaches a board, the tokens have acquired a proposition their signers never made: three experienced IETF participants thought the appellant was correct.

Revision 01 of Lars Eggert's Requiring Support for Appealing to the IESG and IAB anticipates that error. It says a supporter only supports bringing the issue before the entire responsible body and spending that body's time. The support “should in no way” be read as support or non-support for the appellant's view. This is not delicate editorial nuance. It is the constitutional boundary of the proposed mechanism.

The draft responds to an asymmetry. An individual can ask a full review body to examine a dispute, while the collective body bears the work of reading the record, resolving procedure and producing a disposition. The draft reports arguments that appeals have imposed asymmetric workloads and have been used to stall progress. Its proposed answer is social proof of review-worthiness: find three active participants prepared to ask the institution to take the matter up.

That answer remains only a proposal. On 28 August 2026, the Datatracker identified revision 01 as an active individual Internet-Draft. It had no stream, no responsible Area Director, no telechat date and only the IESG state I-D Exists. The Datatracker warns that an individual draft is not endorsed by the IETF and has no formal standing. The text intends Best Current Practice status and says it would update RFC 2026 if approved. The conditional matters. No appellant today should be told that this draft has already rewritten BCP 9.

Where the gate would sit

RFC 2026 starts a Working Group dispute with the chair or chairs. If it cannot be resolved, an involved party may bring it to the relevant Area Director. If that attempt fails, a party may appeal to the IESG as a whole and then, if still dissatisfied, to the IAB. The RFC also provides paths for process failures and claims that the procedures themselves are inadequate. An appeal must describe the facts in detail, begin within two months of public knowledge of the challenged action and receive a disposition within a reasonable period, although each decision body may define its procedure.

The proposed supporter requirement does not close the first doors. A person could still dispute a Working Group recommendation with its chairs and seek resolution from the responsible AD without recruiting supporters. The gate appears when the claim asks the entire IESG or the IAB to take institutional custody of the review.

That placement changes its meaning. It is not a standing rule deciding who may complain. It is not an evidence rule deciding which facts are true. It is not a consensus method deciding which technical position the community prefers. It is a scheduling rule at the point where a dispute consumes the attention of a collegiate appellate body.

The distinction also explains why “three” cannot be treated as a miniature vote. The potential electorate is not defined. No denominator connects three to the IETF community. Supporters do not choose between the appellant and the challenged decision. They do not approve relief. They do not bind the IESG or IAB. Even three hundred such attestations would still leave the review body responsible for the merits; two would not make the facts false.

Qualification imports a map of participation

The draft does not accept any three people. A supporter must be NomCom-eligible under the criteria in RFC 9389, measured as of the date the appeal is raised. The current rules offer three alternative histories: attendance, online or in person, at three of the last five IETF meetings; service as a Working Group chair or secretary within the previous three years; or authorship or editorship of at least two IETF-stream RFCs in the previous five years.

This is an administrable proxy for recent IETF experience. It is not a certificate of expertise in the disputed subject. A person may qualify through transport RFC authorship and be asked to support review of a procedural dispute in another area. Another person may understand the facts intimately yet lack the required participation history. The first can sign the admission request; the second cannot. Neither fact determines who is right.

The imported proxy therefore changes the practical topology of an appeal. A newcomer can begin the dispute but may need access to established participants before a full-body review becomes likely. People who attend meetings, hold formal Working Group roles or publish RFCs become bridges to the appellate queue. That may be a reasonable way to filter unilateral demands. It may also make social reach, not just legal sufficiency, part of admission.

The sources do not establish that valid appeals will be excluded. They do not measure the qualified population, its distribution across regions, languages or technical areas, or the willingness of eligible people to support review without endorsing a position. Those are open empirical questions. The right governance response is not to predict the answer confidently. It is to preserve enough evidence to learn it.

Revision 01 deliberately declines to import RFC 8713 section 4.15's additional NomCom-volunteer exclusions. Members of the IESG and IAB may therefore qualify as supporters. This enlarges the pool, but it creates another distinction to protect. Supporting admission to one's own body is not adjudicating the merits. A transparent recusal or role record may still be prudent when a supporter later participates in the disposition, because readers will otherwise join the two acts themselves.

Time, cooling and a fresh gate upstairs

The mechanics impose more than a count. Three supporters must file with the handling body within two weeks after it receives the appeal. The appellant is responsible for finding them and may send one recruitment message to one public IETF mailing list. Supporters should write to the body personally rather than proxy through the appellant.

Personal filing matters because it gives the handling body a direct attestation. It can distinguish an appellant's claim that someone supports review from that person's own message. It still leaves operational questions. Which timestamp closes the window? What happens to a message delayed by moderation? How is a duplicate or withdrawal treated? What proof establishes eligibility as of the appeal date? How is a mistake corrected without publishing a dossier of meeting history or volunteer status?

The one-year rule adds a moving constraint. A qualified person must not have supported the same appellant in a previous appeal during the preceding year. They may support other appellants. A qualified appellant may support their own appeal, but the draft explains that they can do so only once under the same cooling rule.

This design targets repeat coalitions around one appellant rather than frequent support in general. Four people cannot simply rotate three signatures across an unlimited sequence of appeals by the same member. Yet the rule also makes supporter availability decay with use. A person willing to defend access to review without endorsing merits becomes unavailable to that appellant for a year. If the eligible social neighbourhood is small, the second appeal may encounter a higher practical bar even when it concerns a different act.

Escalation creates another reset. An appellant who goes from the IESG to the IAB must find new qualified supporters. The first three said the IESG should spend time on the matter. They do not automatically say the IAB should spend time reviewing the IESG outcome. That is coherent because each level consumes different institutional attention. It also means the appeal path contains two admission decisions, not one portable mandate.

The number is not absolute

The draft says the handling body may choose not to consider an appeal if it lacks enough qualified supporters. It also says the body may consider the appeal anyway. Three is therefore a presumptive gate with institutional discretion, not a hard jurisdictional limit.

That safety valve is important. A rigid numeric rule could exclude a serious process failure whose facts are visible only to a small or new group. Discretion lets the IESG or IAB hear such a matter. But discretion changes the accountability question. If one two-supporter appeal is heard and another is declined, a bare count cannot explain the difference. The exception needs a reason class: urgent systemic risk, unavailable qualified pool, conflict concentration, material procedural defect, or another bounded rationale.

Unrecorded discretion would produce the worst of both designs. Ordinary appellants would face a visible numeric hurdle, while insiders could appear to receive an invisible exception. A reasoned admission record does not eliminate judgment. It makes clear who exercised it, under which authority and for what procedural reason.

The same discipline protects the review body when it declines. “Only two supporters” should mean two valid attestations under a named rule at a named closing time, not three messages of which one was silently discarded for an eligibility interpretation no reader can inspect. Privacy prevents publication of every underlying record. It does not require publication of nothing.

An admission receipt, not a supporter registry

The smallest useful record is an appeal admission receipt. It should name the appellate level, the filing timestamp and the two-week close. It should bind the exact eligibility rule and its version to the date on which eligibility was assessed. Each support should reach the body personally and carry an explicit semantic assertion: support for review time only; no merits position recorded.

The protected layer can retain the minimum evidence needed to verify eligibility, the one-year same-appellant check, self-support and a fresh set on escalation. The public layer can report aggregate counts: qualified, late, duplicate, withdrawn or ineligible, with bounded reason codes. It need not reveal meeting attendance, volunteer history, private addresses, social relationships or the facts of a sensitive appeal.

The receipt should then record the body's admission decision. If it hears a below-threshold matter, it should publish a non-sensitive reason class. If it declines, it should state the controlling rule, count and correction path. A later merits disposition belongs to a separate field and a separate event. Otherwise a database will eventually turn admitted=true into appeal_supported=true and the original constitutional boundary will disappear.

Correction is part of the design. A supporter may be misclassified. A timestamp may be wrong. A prior support may fall just outside the one-year window. The system should preserve the initial state, correction authority, reason and resulting count instead of overwriting history. Appeals exist partly because institutions can err; their admission record should not assume infallibility.

What the debate should measure

The useful question is not whether three feels high or low. The useful questions are observable. How many potential supporters satisfy each RFC 9389 path at the time of an appeal? How many appeals obtain support through pre-existing relationships rather than public recruitment? How often do eligible people decline because they fear their procedural support will be read as a merits endorsement? How often does the one-year rule remove otherwise willing supporters? How often is discretion used below the threshold, and for which reason classes?

Distribution matters as much as volume. A global count of eligible participants can hide a thin pool in one language, region or technical area. The record should test whether support comes from the appellant's immediate network without turning relationship analysis into public surveillance. Privacy-safe concentration measures and voluntary feedback can reveal structural dependence without publishing a map of dissent.

Every review should also check lexical discipline. Mailing-list discussion already shows how ordinary wording can slide toward “people who think the appeal has merit.” That may be conversational shorthand, but the draft says something narrower. Forms, acknowledgements, dashboards and decisions should all use the same term for admission support. If the interface asks “Do you support this appeal?”, it has already weakened the rule.

Sources