docs(TAP-011): record round failure semantics and ingress validation as open work - #216
Draft
NovusEdge wants to merge 4 commits into
Draft
docs(TAP-011): record round failure semantics and ingress validation as open work#216NovusEdge wants to merge 4 commits into
NovusEdge wants to merge 4 commits into
Conversation
…as open work Follow-up to The-AI-Alliance#165, at @deanwampler's request on merge. Adds three items the merged ADR leaves open, scoped as open work rather than as new decisions: - Durable LinkState persists infrastructure state, not training-round state. Names what a round still has to define (phase checkpoints, mid-merge resume, round identity after a coordinator restart) and defers the design to The-AI-Alliance#183. - The sync FedAvg barrier has no straggler policy, and the substrate default is a silent indefinite wait. Notes that dropping a contribution is governance-adjacent, not a timeout constant. - The mutually-authenticated ingress is asserted but never exercised; the only Tapestry deployment of the transport ran --insecure. Also notes Flower's node-key auth is a different mechanism from the AWS reference's client-certificate mTLS. Adds acceptance criteria 7-9 and the supporting Basis-table rows and references. Signed-off-by: NovusEdge <novusedge0@gmail.com>
deanwampler
requested changes
Aug 6, 2026
deanwampler
left a comment
deanwampler
left a comment
Member
There was a problem hiding this comment.
Thanks for the submission. It is marked "Draft". Is it ready for review and merging?
I think you can drop the references to the WAN spike and M0, which are more transient "states". I.e., you could reword that content to be more forward looking towards our "permanent" solution, which we'll get closer to in M1. These ADRs are intended to have a long horizon viewpoint.
I'll also ask the Flower team to comment where they are mentioned.
Member
|
@niclane7 your feedback is welcome... |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #165, which @deanwampler invited on merge. Writes up the three points from my 07-21 review there against the merged text.
All three land in Consequences and Acceptance criteria as open work. Nothing here touches the sizing call or the governance framing settled on #165.
LinkStatepersists infrastructure state, not training-round state, so it does not answer what a round writes per phase, how a mid-merge failure resumes, or how a node learns its round after a coordinator restart. Design deferred to [EPIC] M1 consortium training #183.--insecure. Also notes Flower's node-key auth is a different mechanism from the AWS reference's client-certificate mTLS.Plus acceptance criteria 7 through 9, the supporting Basis-table rows and references, and a note on criterion 6 about SQLite state on NFS-class storage.