The specification is frozen, so a correction to it is published as an erratum rather than edited into the document. The same discipline applies to MIZAN's own engine and registry — a finding against the implementation is published even when the specification is fine and the implementation is not.
A frozen standard that quietly improves itself is not frozen, and an issuer who edits the text under a published digest has revoked a document without saying so. So the discipline is the same one VTR-1 requires of credentials: weakness is disclosed, not rewritten.
Each erratum records what was found, who found it, and what follows — one of three outcomes:
| Outcome | Meaning |
|---|---|
| no change | The text is defensible as written; the erratum stands as clarification for readers who hit the same question. |
| clarification | The text is unclear or ambiguous but its requirement is unchanged. Recorded here; folded into the next numbered version. |
| requirement change | A requirement genuinely changes. This produces a numbered successor version under §6, which supersedes but never invalidates credentials issued under the prior version. |
Errata are published whether or not the finder wishes to be named, and whether or not the finding flatters the standard. Requirement identifiers R-01 to R-41 are permanent: a retired requirement keeps its number rather than having it reused.
Last updated 1 August 2026 · 4 errata · all resolved · 0 requirement changes · v1.1 published
In Appendix D.2, the first line presents three independent definitions separated by a middle dot:
The · is a separator, but the same glyph denotes multiplication everywhere else in the appendix — including two lines below it, in the PSR denominator. A reader scanning quickly can parse n · m₂ as a product.
Intended reading: three separate definitions — μ = (1/n)Σrᵢ, then m₂ = (1/n)Σ(rᵢ−μ)², then σ = √m₂.
Effect on conformance: none. R-30 is unchanged, and no conformance vector is affected. An implementation that read the line correctly is conforming; one that read it as a product would fail vector 1.1 immediately, so the ambiguity cannot propagate silently into a credential.
Disposition — RESOLVED 1 August 2026. Appendix D was re-set as display mathematics in a typeset edition of v1.0. The ambiguous separator no longer exists: the three definitions are now three separate display equations. No normative text was altered — all 85 normative sentences and all 41 requirement identifiers are verbatim identical between the two renderings, which anyone can confirm by extracting the text of both PDFs. The first rendering remains published and timestamped rather than withdrawn, and both digests are listed on the canonical page.
This is the boundary the freeze actually draws, and it is worth stating plainly: the normative text is frozen; its typography is not. A reader who cannot parse an equation has not been served by a specification that is technically unchanged. What the freeze forbids is altering what the document requires — and that is machine-checkable, so the claim is falsifiable rather than promised.
Found by: internal adversarial review of the document, 1 August 2026, in the same pass that produced §§9–11 and Appendices A–E.
In the first typeset edition (digest 2305c06df6e24b03…), three explanatory paragraphs in E.1, E.2 and E.4 rendered inside the monospace blocks that precede them, rather than as body text. The cause was a paragraph boundary lost while porting the appendix into the print rendering.
Effect on conformance: none. No requirement, identifier or value was affected; the text was present and complete, in the wrong typeface. It is nonetheless a defect — Appendix E exists to be implemented from, and a reader who cannot tell specification from commentary has been poorly served.
Disposition — RESOLVED within the hour. Corrected typeset edition published, digest c68e67f82a1b0aae…. Both superseded renderings remain published and timestamped on the release. Re-verified after the fix: 85/85 normative sentences and 41/41 requirement identifiers verbatim identical to the first rendering.
Found by: full page-by-page visual review of the rendered document — the same review that should have preceded the first publication and did not. Recorded here because a process that only catches the errors it is comfortable admitting is not a process.
The Status block states:
That was true when written, when exactly one rendering of v1.0 existed. It is no longer true. Three renderings are now published — the first rendering, the typeset edition, and the corrected typeset edition — all of them v1.0, all carrying byte-identical normative text, all timestamped, none withdrawn. A copy hashing to a superseded digest is still this version.
Corrected reading: the current rendering's digest, and every superseded rendering's digest, are published on the canonical page. A copy whose digest appears in neither list is not this version.
Effect on conformance: none. No requirement, Level, or verification step is affected. The defect is in a statement about the document's own identity, not about what it requires.
Disposition: the frozen text is not edited. This is normative-section prose, not typography, and the rule this project set for itself is that typography may be re-rendered while text may not. The sentence is corrected in v1.1. Until then this erratum is the authoritative reading.
Found by: page-by-page review of all 22 pages, 1 August 2026. Noted candidly: this defect was introduced by the typeset edition published earlier the same day — a claim about identity broken while fixing typography, which is precisely the class of error this standard exists to catch.
§11 closes "its authors take responsibility for the enforcement" and Appendix C states "The authors cannot close that gap alone." VTR-1 v1.0 has one author. The plural is an editorial slip, not a claim of shared authorship or shared responsibility.
Effect on conformance: none. Disposition: corrected in v1.1; the frozen text stands. Recorded here because a document asserting that claims should be checkable does not get to leave its own inaccuracies unrecorded, however small.
Found by: page-by-page review of all 22 pages, 1 August 2026.
Version 1.1 was published before any citation of 1.0. It resolves Errata 3 and 4 and reworks punctuation and sentence rhythm for readability. No requirement was added, removed, or altered in meaning.
That claim is machine-checkable, which is the only kind of claim this project makes. Extract the Appendix B table from both PDFs and compare: all 41 identifiers carry identical section reference, identical conformance keyword, and identical statement text. A single differing character would falsify it.
Version 1.0 remains published, timestamped, and verifiable. It is superseded, never withdrawn.
All renderings published, all timestamped, none withdrawn. The full list, with the .ots proofs, is on the canonical page.
Non-normative changes to the canonical page. For completeness, since the canonical page is the normative text and readers are entitled to know when anything on it moves: after the freeze, navigation links to this register and to the quant companion were added to the canonical page's footer. No normative text was altered, and the digests published on that page are unchanged and still verify.
The same discipline applies to MIZAN's own engine and registry. Two register entries restate spec Errata 3 and 4 in register form; the rest are findings against the implementation itself.
Before — a valid honest-FAIL credential printed:
After — the same credential now prints:
If you implement VTR-1 and find that something is ambiguous, unreproducible, or wrong, that is more valuable to this standard than a compliment. A specification with one implementation is a description of that implementation.
Send it to [email protected]. Include the requirement identifier if you have one. You will be credited by name unless you ask not to be, and the finding is published either way. A suspected soundness break — a case where a credential verifies that should not — takes priority over everything else and is disclosed as era status under §6: the affected era is marked, never rewritten, and its credentials stay checkable so readers can judge them for themselves.