This push
- pack ICD-AUTH-USER, Q2
- shall GET /users/{id} → 200 or 404
- bytes producer now returns 410
Topic · interface control document
Gist
An interface control document is the INT-REQ that still matches both sides this commit, not a PDF from last quarter. Proof fails the merge with proof audit --check interface_staleness_clean. NASA still writes the signed pack. Jama still authors.
proof audit --check interface_staleness_clean
Keep GPS.gov if you need the signal ICD. Keep NASA 6.3 if the programme wants a signed PDF. A sha256 is not a proof of the function, and a review stamp is not a mechanical drawing.
01 · The PDF that outlived the payload
The merge stayed green because nobody asked whether the bytes the consumer still reads match the contract the producer still ships.
On this install auth still called user. The ICD PDF still said GET /users/{id} returns the profile or 404. The producer changed the body. Reviewers argued about the PDF. There was no INT-REQ whose fingerprint could fail.
proof audit --check interface_coverage fails when a real boundary has no INT-REQ. proof audit --check interface_staleness_clean fails when an implementation artifact drifted from the sha256 stamped on the last interface review. proof review interface INT-REQ-046 --review-rationale "payload shape still matches the consumer" records that review on the requirement itself. It does not write a NASA ICD PDF.
The printed SRS lives on
software requirements specification
and will not write an ICD. The split of one parent into owned children lives on
requirements decomposition.
The caller/callee picture lives on
proof diagram components,
not this H1. This URL is the contract between two components, kept in CI.
proof gaps specs/system --check interface
proof audit --check interface_coverage
proof audit --check interface_staleness_clean
proof review interface INT-REQ-046 --review-rationale "payload shape still matches the consumer"
On this install the teaching miss is a stale pack: the PDF still named the old body, and CI never asked whether the files that implement the call still match the last review.
02 · The exhibit
The last ICD still said 404. The producer now returns 410. Click the tabs.
This push
Finding
No INT-REQ. No fingerprint. Reviewers argued about a PDF the producer no longer implements.
Stale packThis push
Still last week's Go. The check will not invent a contract. Write the INT-REQ, stamp the review, then merge.
Keep the YAMLProof
Same repo. A stale PDF, or a fingerprint from this commit. Click the tabs.
| Invariant | What they do | What Proof does |
|---|---|---|
| Wikipedia ICD | A technical document that records, defines, and controls interface information between two systems. | Not that PDF. Proof keeps an INT-REQ in the repo and fails CI when the implementation fingerprint drifted. |
| GPS / NASA 6.3 | A signed interface pack. gps.gov publishes signal ICDs. NASA 6.3 is interface management, including the approved documentation. | Not a certificate. Not a mechanical drawing. Proof does not sign an ICD for a vehicle or a constellation. |
| Jama / DOORS | An object in the ALM that still names the old payload until someone edits the row. | Jama still authors. IBM DOORS stays a mention on Proof vs Jama. Proof fails the merge from the repo. |
| Component diagram | UML boxes for callers and callees. | proof diagram components --format mermaid prints documented interface edges. Different subcommand. Not this H1. |
| SRS | The shalls for one component, printed from the graph. |
software requirements specification
as proof doc generate npr7150-srs. That command will not write an ICD. Not this H1. |
| A stamp is a proof | A reviewed ICD means both sides still match in production. | A fingerprint match means the files named on the INT-REQ still hash the same. It is not a proof of the function. |
Write the INT-REQ under specs/integration/requirements. A component with no boundary attests no_interface with a reason; inventing a thin contract to silence the check is theater. Stamp a review after the files change. A review without fingerprints cannot detect drift; back-fill with --migrate-fingerprints.
proof gaps specs/system --check interface
proof audit --check interface_coverage
proof review interface INT-REQ-046 --review-rationale "payload shape still matches the consumer"
proof audit --check interface_staleness_clean
We have not run GPS ICD, NASA 6.3, and Proof on the same frozen corpus, and we have not claimed this fingerprint is a signed ICD or a proof of the Go. The loss is named, not scored.
03 · The honest loss
Proof keeps the INT-REQ true in CI. It does not author the programme ICD, qualify a drawing tool, or prove the code. Jama still authors.
The command does not print an SRS. That H1 lives on
software requirements specification.
It does not split a parent into owned children. That H1 lives on
requirements decomposition.
It does not reprint the SpecTree. That H1 lives on
requirements diagram.
There is no ICD template on proof doc generate.
A missing INT-REQ is a spec problem, not a renderer problem. A stamp without fingerprints cannot see drift. Jama still authors.
04 · Nearby questions
What is an interface control document? Wikipedia owns that SERP as the technical document between two systems. Same cluster as this URL only when the contract is an INT-REQ in a repo. Not a twin for a GPS signal ICD.
What is NASA interface management? NASA 6.3. Signed packs and approved drawings. Not this H1. Proof does not replace that process.
What is a component diagram? Callers and callees.
proof diagram components --format mermaid
on this install. Not this H1.
What is a software requirements specification? The shalls for one component.
software requirements specification
as proof doc generate npr7150-srs. That page will not write an ICD.
What is requirements decomposition? One parent split into owned children, including interface children.
requirements decomposition
as proof req decompose.
Is Proof a Jama alternative? No. Jama still authors. Proof fails closed when the implementation hash drifted. IBM DOORS stays a mention on Proof vs Jama.