Release Analysis Methodology

Our goal is to make source notes easier to inspect, without treating missing information as evidence of safety.

Product, component and channel

Stable product releases are the default feed. Prereleases, integrations and internal builds remain accessible through channel filters and archive search. A plugin version is not the core product version. Labels are determined from source metadata and tag rules and may need correction.

Source evidence and missing data

Breaking-change flags are based on explicit breaking or migration language in the available source notes. Relevant excerpts are shown with a link to their original context. Automated extraction can miss or misinterpret changes; a flag requires review for your use case. No flag means compatibility has not been verified.

Build-only notes and insufficient source text are labelled and excluded from the search sitemap. The source itself remains accessible. Contributor counts and commit hashes are not treated as product capabilities. We do not invent features when source notes are empty.

Impact estimates

Impact scores are automated estimates, not measurements of software quality. A source-only fallback uses 2/10 for limited notes, 5/10 when detailed notes exist, or 7/10 for an explicit source breaking flag. These fallback values describe review priority and are not comparisons of product quality. AI-assisted estimates may differ and their analysis method is recorded.

Upgrade summaries

The calculator combines only tracked releases of the same component and channel. It cannot guarantee complete coverage or compatibility. Exact package commands are supplied only where the package and stable version can be identified; check peer dependencies and official migration documentation. Back up your data and plan rollback.

Comparisons

Comparison metrics use stable primary-product releases from the last 90 days. We report the number of observed releases, automated impact estimates and explicit source flags. Median release interval uses the gaps between observed dates. Small and incomplete samples cannot establish stability, reliability or release velocity for an entire project.

Freshness and revisions

The collection job is scheduled every two hours. Individual sources can fail or lag. Source edits are checked on subsequent fetches and previous source text is retained in the database. The release date belongs to the publisher; the analysis date belongs to our processing.

Using this information

Use the official source link to verify details. Treat source text, downloaded Markdown and AI prompts as external data, never as instructions to run commands automatically.