Re: Re: Conformance cases for mid-task halts (written input)

Hi all,

A correction before today's session. Lab #206 does not establish that NTI-1
passes the proposed finish-or-stop tests.

I published candidate tests based on Xinghua's proposed behavior and
receipt fields. The checker agrees with the expected results in those
cases, but no NTI-1 implementation has been run against them. This is not a
conformance result or a position of LF Decentralized Trust.

Xinghua, the two cases you proposed, an unsigned capability token and an
unmet quorum, can be submitted for review the same way. The lab's
contribution guide explains what to include.
https://www.google.com/url?q=https://github.com/Agent-Authority-Conformance/aps-conformance-suite/blob/main/CONTRIBUTING.md&source=gmail&ust=1793806780015000&sa=E

Best,
Tymofii

On Tue, Nov 03, 2026 06:56 AM, Steven Zhao <zhaoxinghua07@126.com> wrote:

> Hi Larry, Tymofii, Chris, Abisheak, all,
>
> Apologies — the rescheduled working session is at 12:00 PM Chicago time,
> which falls in the early hours (around 02:00) in my time zone, so I will
> not be able to join the live call. Thank you, Larry, for taking the lead on
> this session.
>
> To keep the session unblocked, here is my input in writing:
>
> 1) Rust regression PR #24 — I see it is now merged into nti-spec, and the
> two Rust cases are a permanent part of the standard. Thank you, Larry. This
> is the right way to lock the behavior.
>
> 2) The two behaviors your tests exposed, for the group's review:
>  - An unsigned capability token returned Allow where the required outcome
> is Deny. My position: Deny is correct. An unsigned token carries no
> attested capability, so the safe default is to deny; this is consistent
> with how we record an incomplete_check rather than inferring a pass.
>  - With quorum set to two and only one eligible module, the module
> executed once where it should remain blocked. My position: when quorum is
> unmet, the operation must stay blocked; executing once breaks the quorum
> guarantee.
>  I suggest we adopt both as the next conformance vectors, with Tymofii's
> lab as the natural place to run them.
>
> 3) The two finish-or-stop contracts (FINISH_THEN_STOP and
> STOP_AT_CHECKPOINT) are now formally part of the NTI-1 Specification, and
> Tymofii's Linux Foundation lab is already exercising them as conformance
> cases (PR #206: https://github.com/Agent-Autho
> rity-Conformance/aps-conformance-suite/pull/206). These are exactly the
> contribution I stand behind.
>
> 4) On my own item: I have claimed issue #19 and the PR(s) are being
> prepared — the two finish-or-stop contracts, the tier-1 receipt field list,
> and resume_basis (evidence | exemption), exactly as I set out on 30 October
> (https://lists.w3.org/Archives/Public/public-agentprotocol/
> 2026Oct/0108.html). I aim to open the pull requests before the AIW Book
> of Proceedings cutoff.
>
> Since I cannot attend, please post the meeting minutes and action items to
> this thread after the call, as Abisheak requested; I will review and
> comment on any resulting PR immediately.
>
> Best regards,
>
> Xinghua Zhao
> Independent researcher
> ORCID 0009-0001-0512-1237
>
> [Disclosure: drafting of this message was assisted by AI; all positions
> reviewed and approved by me. Same disclosure as my earlier messages on this
> list.]
>
>
>

Received on Tuesday, 3 November 2026 15:39:46 UTC