Now that css containment has dramatically improved large-page
performance, I can crank up the chromium throttling during lighthouse
runs to ensure no perf regressions.
Copy what tantek did by using h-pronoun to link to pronoun.is and
combining p-pronouns (plural) with p-pronoun (singular). Seems to be
supported by some implementations such as Authl.
This also entailed re-writing my "about" page's def-lists in raw HTML,
which was probably long overdue.
FeedValidator was fixed upstream; it doesn't use an allowlist of (X)HTML
attributes anymore. After updating FeedValidator to the latest commit, I
don't need to filter out false-positives for these attributes.
The FeedValidator commit that resolved the issue:
0b130bd5d7
- I don't log IP addresses when you use my Tor hidden service (duh)
- Fix bad timestamp
- Better summary on the top
- Rephrasing
- Mention that webring links do actually send a referring domain
- Replace achecker flags with a config file
- Bring back webhint
- Amend check-whole-site so that it will deploy to staging if all checks
pass, and then run webhint on every staging page.
CSS containmnet crops breadcrumb focus indicators when they overflow
their containers. Instead of adding a new rule for this, refactor some
old rules so padding-increases also apply to the breadcrumb containers.
Add padding to elements so content containment won't cause cropping.
Adjust the global body padding accordingly. This also exposed a
redundancy in the stylesheet, which was removed.
Now the site headers/footers, article elements (including archive
pages), <pre> elements, and other top-level elements are contained.
Rendering long-ass articles with thousands of nodes should be a little
faster.
