# Website Release Diff 1.0.0 Compare two complete static-site ZIP builds and review new, inherited and resolved structural findings. Python 3.10+ standard library. No server connection or deployment. Native JSON editor: `python app.py` (Tkinter). Load input JSON with baseline and candidate ZIP paths, origin, languages and optional protectedIds. Optional baselineRoot/candidateRoot select subfolders inside archives. Prefer absolute ZIP paths. Choose a new .json output; a matching escaped HTML report is written beside it. Open that HTML in any browser to review findings. No scripts or remote resources are included in the report. CLI: `python checker.py input.json new-report.json`. Exit 0 means comparison succeeded, not that it passed: inspect pass/regressions. Exit 2 means invalid input or I/O failure. Neither output may already exist. Run `python examples/make_sample.py` once to create synthetic sample-output ZIPs, then `python checker.py examples/input.json sample-report.json`. Run these commands from the product directory. The example deliberately removes a page. All pages, including unchanged pages, are checked against each complete tree. This catches a removed destination breaking an unchanged inbound link. Checks include missing resources, links/anchors, hreflang targets/reciprocity, removed alternates, canonical metadata, robots changes, lost protected IDs, duplicate titles and removed pages/sitemap URLs. Existing findings are separated so an old defect is not counted as new. Heuristic metadata length rules are editorial defaults, not proof of search rankings. Binary assets are copied only to a temporary private extraction tree and never executed. ZIP hashes bind the comparison. Not an overlay comparison. Archives must represent complete releases. Only UTF-8 .html pages and an optional standard urlset sitemap.xml are supported. No JS rendering, external link checking, sitemap indexes, application routes, CSS parsing or deployment is performed. Robots metadata changes are reviewed; robots.txt interpretation is outside this edition. ZIP admission refuses traversal, Windows special names, links/special files, encrypted entries, duplicate/case-colliding names and excessive expansion. Limits: 100 MiB compressed, 2000 entries, 2 MiB per entry, 50 MiB expanded, 200:1 maximum expansion ratio. Extraction is temporary and removed after processing. Existing site files are never changed. Tests: `python -m unittest discover -s tests -v`. Actual native/HTML visual acceptance is separate from automated source/CLI tests. No publication or platform approval is implied.