spec(0.10.0): §4's probability is p(last | first n−1) among sequences of the entry's own length — "joint prob otherwise" withdrawn

[coderoast-dev]

What

SPEC.md §4 defined behavior.top_ngrams[].probability only in the example's comment: "p(next | prev) for n=2; joint prob otherwise". This PR replaces that comment and adds a paragraph under §4:

An entry's probability is p(last | prefix): its count divided by the summed count of every sequence of the same length the producer counted in the window whose first n − 1 template ids equal the entry's own. Computed before the top_ngrams_size cut. Never a joint probability.

CHANGELOG.md records it under 0.10.0 (still unreleased) › Changed, and the 0.10.0 headline lists it among the breaking items.

Why

• The conditional is what the field is for. §13's ngram_delta.rate_changed compares the value across two windows as a transition rate. A joint value restates frequency, which count already carries, so rate_changed would move with any change in traffic mix, and one field would mean a different quantity at each ngram_size.
• The published population already means it. The reference producer has always emitted the conditional at every order.
• "Same length" is the smallest rule that holds for every sequence a top_ngrams can carry. It also corrects a reference-implementation defect: when top_ngrams held sequences of two lengths, the producer pooled them under one denominator. Only such documents move.

Change class

Breaking by the letter: a producer that followed "joint" becomes non-conformant. Under GOVERNANCE.md §2's 0.x rule (PR #15), the editor merges it with no rfc: issue. No schema change: the field stays a number in [0, 1]. conformance/metalog_validate.py --selftest: 31/31 fixtures pass.

🤖 Generated with Claude Code