You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Successor to #81, which was auto-closed by the merge of #185 (a Closes #81
reference, one second after mergedAt) rather than by anyone deciding it was
done. Several of its items genuinely are done. Eight files are not, and closing
the issue removed the only record of them.
Verified absent from origin/main (14df4227, 2026-07-28)
File
Kind
ARCHITECTURE.adoc
mechanical — an index, not new prose
FAQ.adoc
mechanical
SECURITY-ADVISORIES.adoc
mechanical
.well-known/humans.txt
mechanical
GOVERNANCE.adoc
needs owner decisions
CODEOWNERS-POLICY.adoc
needs owner decisions — and exposes a live violation
.well-known/security.txt
needs one owner decision — and arms a gate
.well-known/ai.txt
needs an owner licence ruling
Two corrections to #81's framing, both worth having on record:
.well-known/ already exists — it holds void.rdf and void.ttl (VoID
dataset descriptors, unrelated). The three text files go alongside them.
MAINTAINERS.adoc already exists at root. Anything claiming to add a MAINTAINERS file is adding a duplicate in the wrong format — and one such
file did in fact appear in an unpushed local sweep commit, which is part of why
this needed re-measuring rather than trusting the issue text.
Already done (so scope stays honest)
EXPLAINME.adoc, CHANGELOG.adoc (single, no dual-CHANGELOG violation), .github/FUNDING.yml, and the crate-doc item closed by #185.
What each mechanical file should contain
ARCHITECTURE.adoc — an index pointing at what already exists
(docs/architecture/abi-ffi.md, docs/architecture/topology.md, docs/vcl-architecture.adoc, docs/VCL-SPEC.adoc, the deployment/drift/
normalisation docs), not a re-telling. Note the architecture decision
records are already written: .machine_readable/6a2/META.a2ml carries
AD-001…AD-007 as decision/rationale/source triples. Transcribe them as a table
and cite the a2ml as canonical. Narrative source is .claude/CLAUDE.md, which
is tracked and public — contrary to Docs max-out vs standards — 8 missing files + 2 violations (Silver→Gold) #81's assumption that it was private.
SECURITY-ADVISORIES.adoc — model on standards/SECURITY-ADVISORIES.adoc
(Closed / Deferred-with-reason, per-advisory exposure, explicit re-evaluate
trigger). Nine advisories are closed; one is open — GHSA-4w2j-m93h-cj5j quinn-proto, high, remote memory exhaustion, 2 alert instances. State plainly
that there are no researcher credits yet rather than omitting the section.
.well-known/humans.txt — team from MAINTAINERS.adoc; stack from the
README structure plus .claude/CLAUDE.md's language policy.
Owner decisions blocking the other four
GOVERNANCE.adoc — TPCF perimeter: does verisimdb sit at P1 (Trusted
Core) or P3? Given formal/*.v Coq proofs and the SPARK/ZKP work, P1 is
arguable, but it is a ruling, not a derivation. Also: succession/bus-factor
clause, and whether the template's 72h / 1-week discussion windows are real
commitments here. Should reference the existing MAINTAINERS.adoc, not
restate it. Note the governance gate checks only README/LICENSE/CONTRIBUTING,
so nothing forces this file's content — it is policy-driven.
CODEOWNERS-POLICY.adoc — writing this while .github/CODEOWNERS stays
as-is would be self-contradictory. It currently has * @hyperpolymathand /.github/ @hyperpolymath plus further self-ping path lines — precisely the
two shapes the estate policy's Rule 1 bans for solo-owned repos, and a known
contributor to notification flooding. Either fix CODEOWNERS in the same
change (recommended) or write down an explicit exemption.
.well-known/security.txt — the only open field is Expires:. Be aware
this arms a gate: the governance wellknown job currently only warns
that the file is missing, but once it exists it hard-errors on a missing or
past Expires. So adding this file schedules a future red. Suggest 2027-07-31, plus a renewal reminder and an inline comment so the next
maintainer knows the eventual failure is by design.
.well-known/ai.txt — the available estate model cites a Palimpsest/MIT
stance, but verisimdb is MPL-2.0 code + CC-BY-SA-4.0 prose, and META.a2ml AD-006 records "MPL-2.0 across the workspace (no AGPL)". Ruling
needed: adopt Palimpsest conditional-training terms, or a plain
MPL/CC-BY-SA attribution stance? Do not copy the model file unexamined.
Constraints for whoever writes these
Never write the string VeriSim Query Language.tests/doc-consonance-gate.sh
fails the build on it in any *.adoc/*.md/*.a2ml. Write VeriSim
Consonance Language (VCL).
House style: // SPDX-License-Identifier: CC-BY-SA-4.0 + copyright, = Title, :toc: left. REUSE.toml already globs **/*.adoc → CC-BY-SA-4.0 and .well-known/** → MPL-2.0, so the .txt files take MPL-2.0 — don't fight it.
reuse lint must stay compliant (776/776 on main today, +8 with these).
Successor to #81, which was auto-closed by the merge of #185 (a
Closes #81reference, one second after
mergedAt) rather than by anyone deciding it wasdone. Several of its items genuinely are done. Eight files are not, and closing
the issue removed the only record of them.
Verified absent from
origin/main(14df4227, 2026-07-28)ARCHITECTURE.adocFAQ.adocSECURITY-ADVISORIES.adoc.well-known/humans.txtGOVERNANCE.adocCODEOWNERS-POLICY.adoc.well-known/security.txt.well-known/ai.txtTwo corrections to #81's framing, both worth having on record:
.well-known/already exists — it holdsvoid.rdfandvoid.ttl(VoIDdataset descriptors, unrelated). The three text files go alongside them.
MAINTAINERS.adocalready exists at root. Anything claiming to add aMAINTAINERSfile is adding a duplicate in the wrong format — and one suchfile did in fact appear in an unpushed local sweep commit, which is part of why
this needed re-measuring rather than trusting the issue text.
Already done (so scope stays honest)
EXPLAINME.adoc,CHANGELOG.adoc(single, no dual-CHANGELOG violation),.github/FUNDING.yml, and the crate-doc item closed by #185.What each mechanical file should contain
ARCHITECTURE.adoc— an index pointing at what already exists(
docs/architecture/abi-ffi.md,docs/architecture/topology.md,docs/vcl-architecture.adoc,docs/VCL-SPEC.adoc, the deployment/drift/normalisation docs), not a re-telling. Note the architecture decision
records are already written:
.machine_readable/6a2/META.a2mlcarriesAD-001…AD-007 as decision/rationale/source triples. Transcribe them as a table
and cite the a2ml as canonical. Narrative source is
.claude/CLAUDE.md, whichis tracked and public — contrary to Docs max-out vs standards — 8 missing files + 2 violations (Silver→Gold) #81's assumption that it was private.
FAQ.adoc— answers already exist, scattered:EXPLAINME.adoc,README's "How It Compares",
docs/vcl-vs-sql.adoc,KNOWN-ISSUES.adoc,docs/business/pr/faq.adoc. Harvest, don't re-answer.SECURITY-ADVISORIES.adoc— model onstandards/SECURITY-ADVISORIES.adoc(Closed / Deferred-with-reason, per-advisory exposure, explicit re-evaluate
trigger). Nine advisories are closed; one is open — GHSA-4w2j-m93h-cj5j
quinn-proto, high, remote memory exhaustion, 2 alert instances. State plainlythat there are no researcher credits yet rather than omitting the section.
.well-known/humans.txt— team fromMAINTAINERS.adoc; stack from theREADME structure plus
.claude/CLAUDE.md's language policy.Owner decisions blocking the other four
GOVERNANCE.adoc— TPCF perimeter: does verisimdb sit at P1 (TrustedCore) or P3? Given
formal/*.vCoq proofs and the SPARK/ZKP work, P1 isarguable, but it is a ruling, not a derivation. Also: succession/bus-factor
clause, and whether the template's 72h / 1-week discussion windows are real
commitments here. Should reference the existing
MAINTAINERS.adoc, notrestate it. Note the governance gate checks only README/LICENSE/CONTRIBUTING,
so nothing forces this file's content — it is policy-driven.
CODEOWNERS-POLICY.adoc— writing this while.github/CODEOWNERSstaysas-is would be self-contradictory. It currently has
* @hyperpolymathand/.github/ @hyperpolymathplus further self-ping path lines — precisely thetwo shapes the estate policy's Rule 1 bans for solo-owned repos, and a known
contributor to notification flooding. Either fix CODEOWNERS in the same
change (recommended) or write down an explicit exemption.
.well-known/security.txt— the only open field isExpires:. Be awarethis arms a gate: the governance
wellknownjob currently only warnsthat the file is missing, but once it exists it hard-errors on a missing or
past
Expires. So adding this file schedules a future red. Suggest2027-07-31, plus a renewal reminder and an inline comment so the nextmaintainer knows the eventual failure is by design.
.well-known/ai.txt— the available estate model cites a Palimpsest/MITstance, but verisimdb is MPL-2.0 code + CC-BY-SA-4.0 prose, and
META.a2mlAD-006 records "MPL-2.0 across the workspace (no AGPL)". Rulingneeded: adopt Palimpsest conditional-training terms, or a plain
MPL/CC-BY-SA attribution stance? Do not copy the model file unexamined.
Constraints for whoever writes these
VeriSim Query Language.tests/doc-consonance-gate.shfails the build on it in any
*.adoc/*.md/*.a2ml. Write VeriSimConsonance Language (VCL).
// SPDX-License-Identifier: CC-BY-SA-4.0+ copyright,= Title,:toc: left.REUSE.tomlalready globs**/*.adoc→ CC-BY-SA-4.0 and.well-known/**→ MPL-2.0, so the.txtfiles take MPL-2.0 — don't fight it.reuse lintmust stay compliant (776/776 on main today, +8 with these).