CVS witnesses execution and produces independently verifiable evidence.
It does not define what is admissible. It does not enforce admissibility. Those functions belong to the upstream layers.
| Layer | Function |
|---|---|
| Constraint Architecture | Defines what is admissible — the boundaries within which execution is permitted |
| 512 | Enforces admissibility at the commit boundary — binary, deterministic, at machine speed |
| CVS | Witnesses execution and produces independently verifiable evidence |
Constraint Architecture defines the world. 512 enforces it. CVS proves it.
CVS observes execution outcomes against declared constraints. It does not define those constraints. It does not validate whether those constraints are correctly formed. It does not enforce them.
The formation of constraint boundaries — consent logic, authority models, thresholds, domain-specific admissibility rules — is the responsibility of the Constraint Architecture layer.
Errors in constraint definition are upstream failures, not CVS failures.
CVS will faithfully record adherence to incorrectly defined constraints. That is by design. It is also the point: CVS proves what happened against the declared boundary, regardless of whether that boundary was correctly designed.
This property is what makes CVS evidence independently defensible. The evidence does not depend on the correctness of the constraints — only on whether the declared constraints were satisfied.
Constraint Architecture is documented at:
https://github.com/JonathanMastersWatson/Constraint-Architecture
If you are implementing CVS, the constraint boundary is an input, not something CVS produces or validates.
Before instrumenting with CVS, the deploying organisation must answer:
What is the declared constraint boundary this system operates within?
That question is answered upstream. CVS proves adherence to the answer. It does not produce or validate the answer.
If no constraint boundary is declared, CVS has nothing to prove adherence to. Deploying CVS without a declared constraint boundary produces evidence of execution without context — not evidence of legitimate execution.