Skip to content

C-63 ruled — readers do not refuse a newer edition; accepted risk with the argument recorded - #126

Merged
Polichinel merged 1 commit into
developmentfrom
docs/c63-ruled-accepted
Aug 13, 2026
Merged

C-63 ruled — readers do not refuse a newer edition; accepted risk with the argument recorded#126
Polichinel merged 1 commit into
developmentfrom
docs/c63-ruled-accepted

Conversation

@Polichinel

Copy link
Copy Markdown
Contributor

Operator ruling on #45. Register-only: no contract change, no registry change, no version bump — the ruling is that nothing should be added.

The ruling: no

Not because the exposure is small, but because the check would be worse than it:

  1. Covered from the other side. §10.1 tells a consumer when an edition obliges it. A reader refusing on version fires on every edition — including the seven that oblige nobody, which is the noise Let consumers pin the contract version, and mark observation-only bumps as non-blocking #76 was filed to remove.
  2. A fourth pin per consumer, unchecked. Each already pins a tag or version. A second "written against" number must agree with the first and nothing would check that it does — C-53's shape.
  3. The live instance is not a version problem. D-05 diverges on the same edition.

Why the exposure shrank on its own

The registry grew in tables the readers structurally ignore — they scan exactly [connection] and [target], so [contract] and [edition] were invisible by construction. What remains is a change within a scanned table, of which D-05 is the only instance.

Offered, not imposed

The readers' ignore-list is implicit and its comment is three tables out of date. A reader that fails on an unknown table catches the class for one line each — the guard this repo shipped as C-75. A pattern for views-models, views-faoapi and views-crafdapi; no obligation, nothing to re-pin.

Stays in the register

Nothing was fixed — a risk was measured, argued and accepted. Retained so the next person meets the argument rather than the question, with the voiding condition stated: if a future edition alters [connection] or [target] semantics, this ruling is void.

Gates: ruff clean · validate_docs.sh green · pytest 10 red / 24 passed / 2 xfailed.

Closes #45

🤖 Generated with Claude Code

…n; accepted risk

Operator ruling 2026-08-13, views-appwrite#45 closed. No contract change, no registry
change, no version bump -- the ruling is that nothing should be added.

THE RULING IS THAT THE CHECK WOULD BE WORSE THAN THE EXPOSURE, not that the exposure is
negligible:

  1. Covered from the other side. obliges_consumers_since + §10.1 tell a consumer when
     an edition asks something of it. A reader refusing on VERSION fires on every
     edition, including the seven that oblige nobody -- the noise #76 was filed to
     remove. A gate that cries wolf is bypassed reflexively, and that is how its one
     real firing goes unread.
  2. A fourth pin per consumer, unchecked. Each already carries a pinned tag or version
     constant. A second "written against" number must agree with the first and NOTHING
     WOULD CHECK THAT IT DOES -- C-53's exact shape.
  3. The live instance is not a version problem. D-05's divergence happens on the SAME
     edition; a version gate would not have caught it.

WHY THE EXPOSURE SHRANK WITHOUT ANYONE ACTING. The registry's growth has been in tables
the readers structurally ignore -- they scan exactly [connection] and [target], so
[contract] and [edition] were invisible by construction. What remains is narrower than
filed: a change WITHIN a scanned table that an old reader misreads, of which D-05 is the
only instance.

THE CHEAPER ALTERNATIVE, offered and not imposed. The readers' ignore-list is implicit,
and its comment is already three tables out of date. A reader that fails on a table it
has never heard of catches the class for one line each. This repo shipped exactly that
guard on its own side (C-75, mutation-proven both directions). A pattern, not an
obligation: no clause, nothing to re-pin.

Two of #45's three questions were already answered before the ruling, by work that did
not know it was answering them -- major-vs-minor superseded by §10.1, and the constant's
location settled by practice in three repos.

STAYS IN THE REGISTER rather than moving to Resolved. Nothing was fixed: a risk was
measured, argued and accepted. Retained so the next person meets the ARGUMENT rather
than the question -- and with the condition that voids it stated: if a future edition
alters [connection] or [target] semantics, this ruling is void and the question returns.

Gates: ruff clean, validate_docs green, pytest 10 red / 24 passed / 2 xfailed.

Closes #45

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@Polichinel
Polichinel merged commit 601ec86 into development Aug 13, 2026
4 checks passed
@Polichinel
Polichinel deleted the docs/c63-ruled-accepted branch August 13, 2026 03:12
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.

No registry reader checks [meta] version — D-05 is one instance of a class with no policy

1 participant