seo: three pages shared one description and none had a canonical

[ChelseaKR]

What was wrong

render.py hardcoded one sentence into the head of all three pages:

Coverage and completeness counts over California's public wildfire datasets. Unofficial.

So a search result for the DINS page and a search result for the perimeters page were indistinguishable, and neither said what its own page counts. There was also no canonical and no Open Graph on any page.

What this changes

page() now takes a description per page, each restating what that page already says of itself in its own standfirst, plus a canonical, og:* and twitter:card.

page	canonical
index.html	https://chelseakr.github.io/perimeter/
perimeters.html	https://chelseakr.github.io/perimeter/perimeters.html
dins.html	https://chelseakr.github.io/perimeter/dins.html

SITE_URL and CANONICAL_PATH are keyed by active, the value each page already uses to mark its own nav entry, so the canonical and the highlighted tab cannot disagree about which page this is.

The descriptions state no count

Deliberately. The committed site/ is built from data/raw/; the test build is built from the fixtures. A figure hardcoded in render.py would be true of one build and false of the other. CONTRIBUTING's rule is that a number is counted or it is not published, and these counts belong to CAL FIRE's files, which change underneath us. Nothing added here says how many records, how many incidents, or how much is covered. No og:image either: this repository ships no image, and twitter:card is summary, which promises none.

Why a missing canonical matters more here

These pages are one of six project sites on the shared chelseakr.github.io origin, served from paths rather than domains of their own. A canonical of / is not this site's root; it is a different address, it 404s, and all six sites would claim it. test_no_published_link_is_rooted_at_the_server already holds that property for href but cannot catch a canonical, because a canonical is an absolute URL rather than a rooted path.

The gate

The parser in tests/test_pages_html.py recorded name= metas only, so every property="og:..." tag was invisible to it and <link rel="canonical"> is not a meta tag at all. It now records both. The existing title-and-description checks were extended, not joined by a parallel SEO suite. tests/test_published_site.py gains the same assertion over the committed bytes.

break	failure
canonical line deleted	index.html has no canonical URL
SITE_URL set to the bare origin	index.html points at the shared origin, which is a different site
CANONICAL_PATH collapsed to one URL	assert 1 == 3
the original defect restored, one shared description	assert 1 == 3
committed site/index.html canonical hand-edited to the origin	index.html canonicalises to 'https://chelseakr.github.io/'

Restored, make verify exits 0, including html-validate, axe, the Playwright reflow specs and the determinism check.

make site was run from the real data

• A make site on the unmodified tree first produced an empty diff, so the committed figures already matched data/raw/.
• After the change, the site/ diff is head-only: every changed line is a canonical, og:*, twitter:card or description tag. Not one published figure moved.

Not changed

No repository setting, Pages config, ruleset or DNS. No sitemap.xml or robots.txt was added: robots.txt under a project path is inert, because only the origin root is honoured and https://chelseakr.github.io/robots.txt 404s.

🤖 Generated with Claude Code