Skip to content

Prepare protocol v0.4.0 response framing migration - #31

Draft
nirmalhk7 wants to merge 2 commits into
NTARI-RAND:mainfrom
nirmalhk7:nirmal/protocol-response-framing
Draft

nirmalhk7 wants to merge 2 commits into
NTARI-RAND:mainfrom
nirmalhk7:nirmal/protocol-response-framing

Conversation

@nirmalhk7

@nirmalhk7 nirmalhk7 commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

This draft prepares SoHoLINK to consume sohocloud-protocol v0.4.0 from protocol PR #16. The release remains unpublished. Real module checksums, clean tagged-module verification, and coordinated migration review remain required; this PR is not ready to merge or deploy.

The dependency pin moves from v0.1.0 to v0.4.0. The real HTTP regression serves a signed fee declaration through SoHoLINK's actual protocol handler: whitespace preserves the signature, while an appended second document is rejected by the strict reference client. Handler authentication, fee terms, and deployment-wide response-body policy are unchanged.

Current main has been merged, preserving its response limits, uptime scoring, preflight, and job-history changes. The broken lint configuration now validates against golangci-lint v2: supported static checks remain enabled and gofmt uses the formatter section. The unsupported deadcode linter is removed; whole-program reachability uses the separate Go deadcode tool, not unused. The conformance note records this distinction and retains the release gate.

Verification refreshed November 17, 2026, at c3f8012 using Go 1.26.10:

  • Against protocol candidate d0acd07e56e12c3c95ac98877f040ae3ca79fb00 through an external temporary Go workspace, the full race-enabled database integration suite (go test -race -tags integration -count=1 -p 1 ./...) passes against a disposable TimescaleDB database. Required portal/agent/network unit tests, focused signed-fee framing regression, go build ./..., and go vet ./... pass.
  • The same framing regression fails against actual v0.1.0 with “reference client accepted a second response value.”
  • golangci-lint config verify passes. Complete uncapped lint runs report the same 121 findings on current main and candidate, with zero additions or removals. The separate deadcode runs report identical sets of seven findings. Existing findings remain; this is not a clean-lint claim.
  • git diff --check origin/main...HEAD passes. The repair commit carries a DCO sign-off.
  • With the workspace disabled, v0.4.0 download fails with unknown revision v0.4.0; the tag lookup also returns not found. Candidate-workspace validation does not satisfy published-release verification. No committed replacement or fabricated checksum is included.

The full upgrade is not wire-compatible with v0.1.0. Prior GPU/storage additions changed listing and assignment signing bytes under existing v0 domains, including when new fields are empty. Before deployment, agree and verify a coordinated agent, frontend/relay, and coordinator rollout. This draft adopts no persisted-signature migration policy.

After maintainers merge protocol #16 and publish its immutable v0.4.0 tag, obtain the real checksums and rerun checks without the workspace. Hosted CI and independent human approval remain separate requirements. Contributions follow NTARI's Bug-Build-Test guide; the author must not merge the PR.

Pin the proposed protocol release and add a real HTTP response-framing
regression. Document unpublished-tag verification and the coordinated
upgrade required for legacy listing and assignment signatures.

Signed-off-by: Nirmal Khedkar <nirmalhk7@gmail.com>
Merge current main and preserve its response limits, uptime scoring, preflight, and job-history behavior. Migrate lint settings to the v2 schema, retain supported checks and formatting, and document the separate whole-program reachability check. Keep release and migration gates explicit.

Signed-off-by: Nirmal Khedkar <nirmalhk7@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant