ChelseaKR · ChelseaKR · Aug 29, 2026 · Aug 29, 2026
@@ -6,6 +6,44 @@ Format: [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); this project a
 
 ## [Unreleased]
 
+### Fixed, the three pages could not say which of them a search result was for
+
+- **One meta description was shared by all three pages, and none of them had a canonical.**
+  `render.py` hardcoded a single sentence, "Coverage and completeness counts over
+  California's public wildfire datasets. Unofficial.", into the head of the index, the
+  perimeters page and the DINS page alike, so a search result for one was
+  indistinguishable from a search result for another and none said what its own page
+  counts. `page()` now takes a `description` per page, and each restates what that page
+  already says of itself in its own standfirst.
+
+  The descriptions state **no count**. The committed `site/` is built from `data/raw/`
+  and the test build is built from the fixtures, so a figure hardcoded in `render.py`
+  would be true of one 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.
+
+- **A canonical URL and Open Graph tags on every page.** There were none. These pages are
+  one of six project sites on the shared `chelseakr.github.io` origin, served from paths
+  rather than domains of their own, so the canonical a single-domain habit produces (`/`)
+  is not this site's root but a different address that 404s, and all six would claim it.
+  `SITE_URL` and `CANONICAL_PATH` are keyed by `active`, the value each page already uses
+  to mark its own nav entry, so a canonical and the highlighted tab cannot disagree about
+  which page this is. No `og:image`: this repository ships no image, and `twitter:card` is
+  `summary`, which promises none.
+
+- **The page parser in `tests/test_pages_html.py` could not see any of this.** It 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, and the existing
+  title-and-description checks were extended rather than joined by a parallel SEO suite.
+  `tests/test_published_site.py` gains the same assertion over the committed bytes, beside
+  the rooted-link check that holds the same subpath property for `href`.
+
+  Observed failing four ways: the canonical line deleted; `SITE_URL` set to the bare
+  origin; `CANONICAL_PATH` collapsed so all three pages share one canonical; and the
+  original defect restored, one description for all three pages. A fifth, hand-editing the
+  committed `site/index.html` canonical to the shared origin, fails the published-bytes
+  check. `make site` reproduces `site/` from `data/raw/` with no change outside the head.
+
 ### Documented, the pages have been served since 2026-08-08 and the README never said so
 
 - **The README named the two measurement pages only as build outputs.** `site/perimeters.html`

@@ -4,7 +4,15 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>DINS damage inspection coverage | Perimeter</title>
-<meta name="description" content="Coverage and completeness counts over California's public wildfire datasets. Unofficial.">
+<meta name="description" content="What California&#x27;s post-fire damage inspection records contain, counted per field and per incident. Unofficial.">
+<link rel="canonical" href="https://chelseakr.github.io/perimeter/dins.html">
+<meta property="og:type" content="website">
+<meta property="og:site_name" content="Perimeter">
+<meta property="og:title" content="DINS damage inspection coverage | Perimeter">
+<meta property="og:description" content="What California&#x27;s post-fire damage inspection records contain, counted per field and per incident. Unofficial.">
+<meta property="og:url" content="https://chelseakr.github.io/perimeter/dins.html">
+<meta property="og:locale" content="en_US">
+<meta name="twitter:card" content="summary">
 <style>:root {
   color-scheme: light;
   --surface: #fcfcfb;

@@ -4,7 +4,15 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>Perimeter | Coverage measures for California wildfire datasets</title>
-<meta name="description" content="Coverage and completeness counts over California's public wildfire datasets. Unofficial.">
+<meta name="description" content="Two coverage measurements over California&#x27;s public wildfire datasets, published as counts. Unofficial.">
+<link rel="canonical" href="https://chelseakr.github.io/perimeter/">
+<meta property="og:type" content="website">
+<meta property="og:site_name" content="Perimeter">
+<meta property="og:title" content="Perimeter | Coverage measures for California wildfire datasets">
+<meta property="og:description" content="Two coverage measurements over California&#x27;s public wildfire datasets, published as counts. Unofficial.">
+<meta property="og:url" content="https://chelseakr.github.io/perimeter/">
+<meta property="og:locale" content="en_US">
+<meta name="twitter:card" content="summary">
 <style>:root {
   color-scheme: light;
   --surface: #fcfcfb;

@@ -4,7 +4,15 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>Historical fire perimeter completeness | Perimeter</title>
-<meta name="description" content="Coverage and completeness counts over California's public wildfire datasets. Unofficial.">
+<meta name="description" content="How complete California&#x27;s historical fire perimeter record is, counted per fire year and per field. Unofficial.">
+<link rel="canonical" href="https://chelseakr.github.io/perimeter/perimeters.html">
+<meta property="og:type" content="website">
+<meta property="og:site_name" content="Perimeter">
+<meta property="og:title" content="Historical fire perimeter completeness | Perimeter">
+<meta property="og:description" content="How complete California&#x27;s historical fire perimeter record is, counted per fire year and per field. Unofficial.">
+<meta property="og:url" content="https://chelseakr.github.io/perimeter/perimeters.html">
+<meta property="og:locale" content="en_US">
+<meta name="twitter:card" content="summary">
 <style>:root {
   color-scheme: light;
   --surface: #fcfcfb;

@@ -24,6 +24,18 @@
 from perimeter.schema import Basis
 from perimeter.sources import DINS, FRAP, Source
 
+# Where these pages are served. The path segment is load-bearing: this is one of six project
+# sites on the shared `chelseakr.github.io` origin, published on PATHS rather than on domains
+# of their own. `https://chelseakr.github.io/` is not a shorter spelling of this site's root;
+# it is a different address, it 404s, and all six sites would claim it. A canonical naming it
+# would invite a crawler to fold six unrelated projects into one document.
+SITE_URL = "https://chelseakr.github.io/perimeter/"
+
+# Keyed by `active`, the value each page already uses to mark its own nav entry, so a
+# canonical and the highlighted tab cannot disagree about which page this is. The index
+# canonicalises to the directory form, which is the URL Pages serves it at.
+CANONICAL_PATH = {"index": "", "perimeters": "perimeters.html", "dins": "dins.html"}
+
 DISCLAIMER = (
     "Unofficial. Not affiliated with or endorsed by CAL FIRE, FRAP, or any California "
     "state agency."
@@ -437,6 +449,7 @@ def field_table(
 def page(
     *,
     title: str,
+    description: str,
     body: str,
     active: str,
     is_fixture: bool,
@@ -457,13 +470,22 @@ def page(
         for href, label, key in nav_items
         for current in [' aria-current="page"' if key == active else ""]
     )
+    canonical = SITE_URL + CANONICAL_PATH[active]
     return f"""<!doctype html>
 <html lang="en">
 <head>
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>{esc(title)}</title>
-<meta name="description" content="Coverage and completeness counts over California's public wildfire datasets. Unofficial.">
+<meta name="description" content="{esc(description)}">
+<link rel="canonical" href="{esc(canonical)}">
+<meta property="og:type" content="website">
+<meta property="og:site_name" content="Perimeter">
+<meta property="og:title" content="{esc(title)}">
+<meta property="og:description" content="{esc(description)}">
+<meta property="og:url" content="{esc(canonical)}">
+<meta property="og:locale" content="en_US">
+<meta name="twitter:card" content="summary">
 <style>{STYLESHEET}</style>
 </head>
 <body>
@@ -680,6 +702,10 @@ def perimeters_page(report: PerimeterReport, *, is_fixture: bool) -> str:
 """
     return page(
         title="Historical fire perimeter completeness | Perimeter",
+        description=(
+            "How complete California's historical fire perimeter record is, counted "
+            "per fire year and per field. Unofficial."
+        ),
         body=body,
         active="perimeters",
         is_fixture=is_fixture,
@@ -913,6 +939,10 @@ def dins_page(report: DinsReport, *, is_fixture: bool) -> str:
 """
     return page(
         title="DINS damage inspection coverage | Perimeter",
+        description=(
+            "What California's post-fire damage inspection records contain, counted "
+            "per field and per incident. Unofficial."
+        ),
         body=body,
         active="dins",
         is_fixture=is_fixture,
@@ -1005,6 +1035,10 @@ def index_page(
 """
     return page(
         title="Perimeter | Coverage measures for California wildfire datasets",
+        description=(
+            "Two coverage measurements over California's public wildfire datasets, "
+            "published as counts. Unofficial."
+        ),
         body=body,
         active="index",
         is_fixture=is_fixture,

@@ -34,7 +34,7 @@
 
 from perimeter.cells import present_tenths_of_percent
 from perimeter.cli import build_site
-from perimeter.render import DARK, LIGHT
+from perimeter.render import DARK, LIGHT, SITE_URL
 from perimeter.sources import SOURCES
 
 ROOT = Path(__file__).resolve().parents[1]
@@ -75,6 +75,11 @@ def __init__(self) -> None:
         self.lang: str | None = None
         self.title: str = ""
         self.metas: dict[str, str] = {}
+        # Open Graph uses `property=`, not `name=`, so the meta collector above cannot see
+        # it, and `<link rel="canonical">` is not a meta tag at all. Both are head facts a
+        # structural check should be able to reach.
+        self.properties: dict[str, str] = {}
+        self.links: dict[str, str] = {}
         self.prose: list[str] = []
         self.cells: list[str] = []
         self._heading: str | None = None
@@ -106,6 +111,10 @@ def _note_head(self, tag: str, attr: dict[str, str]) -> None:
             key = attr.get("name") or ("charset" if "charset" in attr else "")
             if key:
                 self.metas[key] = attr.get("content", attr.get("charset", ""))
+            if "property" in attr:
+                self.properties[attr["property"]] = attr.get("content", "")
+        elif tag == "link" and "rel" in attr:
+            self.links[attr["rel"]] = attr.get("href", "")
 
     def _note_table(self, tag: str, attr: dict[str, str]) -> None:
         if tag == "table":
@@ -173,6 +182,81 @@ def test_the_page_has_a_title_and_a_description(built: Path, name: str) -> None:
     assert doc.metas.get("description", "").strip()
 
 
+def test_each_page_describes_itself_rather_than_the_other_two(built: Path) -> None:
+    """One description shared by three pages describes none of them.
+
+    All three pages carried the same sentence, 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. The renderer now takes the description per page.
+
+    The descriptions state no count on purpose. The committed ``site/`` is built from
+    ``data/raw/`` and this test builds from the fixtures, so a figure hardcoded in
+    ``render.py`` would be true of one build and false of the other, and CONTRIBUTING's
+    rule is that a number is counted or it is not published. The counts also belong to
+    CAL FIRE's files, which change underneath us.
+    """
+    descriptions = {
+        name: parse(built / name).metas.get("description", "").strip() for name in PAGES
+    }
+    assert all(descriptions.values()), descriptions
+    assert len(set(descriptions.values())) == len(PAGES), descriptions
+
+
+@pytest.mark.parametrize("name", PAGES)
+def test_the_page_names_itself_and_not_the_shared_origin(
+    built: Path, name: str
+) -> None:
+    """The canonical carries this project's path segment, not the bare origin.
+
+    These pages are one of six project sites on ``chelseakr.github.io``, served from
+    PATHS rather than from domains of their own. That makes the single-domain habit
+    actively wrong here rather than merely untidy: a canonical of "/" resolves to
+    ``https://chelseakr.github.io/``, which is not this site's root but a different
+    address that 404s, and all six sites would claim the identical canonical. A crawler
+    that believes them folds six unrelated projects into one document.
+
+    So this asserts the path segment is PRESENT, not merely that a canonical exists: an
+    empty or origin-rooted canonical satisfies "is there a canonical" and is the bug.
+    """
+    doc = parse(built / name)
+
+    canonical = doc.links.get("canonical", "")
+    assert canonical, f"{name} has no canonical URL"
+    assert canonical.startswith(SITE_URL), f"{name} canonicalises to {canonical!r}"
+    assert canonical.rstrip("/") != "https://chelseakr.github.io", (
+        f"{name} points at the shared origin, which is a different site"
+    )
+    assert "/perimeter/" in canonical, f"{name}: {canonical!r} omits the project path"
+
+    # The social card repeats the page rather than describing it a second time, so the two
+    # cannot drift into disagreeing about what this page is.
+    assert doc.properties.get("og:url") == canonical, f"{name} og:url disagrees"
+    assert doc.properties.get("og:title") == doc.title.strip(), (
+        f"{name} og:title disagrees"
+    )
+    assert doc.properties.get("og:description") == doc.metas.get("description"), name
+    assert doc.properties.get("og:type") == "website", name
+
+    # No image is published in this repository, so the card must not promise one. If a
+    # social image is ever committed, this fails until og:image is added with it.
+    card = doc.metas.get("twitter:card")
+    assert card in ("summary", "summary_large_image"), (
+        f"{name} twitter:card is {card!r}"
+    )
+    if card == "summary_large_image":
+        assert doc.properties.get("og:image"), (
+            f"{name} promises an image it has none of"
+        )
+
+
+def test_the_three_pages_do_not_share_one_canonical(built: Path) -> None:
+    """Three pages, three canonicals. A shared one asks a crawler to drop two of them."""
+    canonicals = {
+        name: parse(built / name).links.get("canonical", "") for name in PAGES
+    }
+    assert len(set(canonicals.values())) == len(PAGES), canonicals
+
+
 @pytest.mark.parametrize("name", PAGES)
 def test_there_is_exactly_one_first_level_heading(built: Path, name: str) -> None:
     levels = [tag for tag, _ in parse(built / name).headings]

@@ -121,6 +121,37 @@ def test_no_published_link_is_rooted_at_the_server(name: str) -> None:
     )
 
 
+@pytest.mark.parametrize("name", PAGES)
+def test_every_published_page_canonicalises_to_itself_under_the_subpath(
+    name: str,
+) -> None:
+    """The same subpath property the test above holds for links, held for the canonical.
+
+    A link rooted at ``/`` leaves this site. So does a canonical: ``site/`` is served
+    from ``/perimeter/`` on an origin shared with five other project sites, so a
+    canonical of ``https://chelseakr.github.io/`` names a different address, one that
+    404s, and every one of the six would claim it. ``ROOTED`` above cannot catch it,
+    because a canonical is an absolute URL rather than a rooted path.
+
+    This checks the COMMITTED bytes. ``tests/test_pages_html.py`` holds the same property
+    over a fixture build, which proves the renderer; only this proves what is deployed.
+    """
+    expected = {
+        "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",
+    }[name]
+    text = (SITE / name).read_text(encoding="utf-8")
+    found = re.search(r'<link rel="canonical" href="([^"]*)">', text)
+    assert found, f"{name} has no canonical URL"
+    assert found.group(1) == expected, (
+        f"{name} canonicalises to {found.group(1)!r}, not {expected!r}"
+    )
+    assert f'<meta property="og:url" content="{expected}">' in text, (
+        f"{name} og:url disagrees with its canonical"
+    )
+
+
 @pytest.mark.parametrize("name", PAGES)
 def test_every_published_page_says_it_is_not_affiliated_with_cal_fire(
     name: str,