Accessibility

This archive is built as plain, semantic HTML: real headings, real landmarks, real links. It works without JavaScript, and every page is readable at any window width.

What we have checked

Every page declares its language and uses header, navigation, main and footer landmarks. Every page carries a skip-to-content link as its first element, and interactive elements show a visible focus ring when you tab to them.

Contrast ratios below were computed from the shipped :root tokens against the WCAG 2.1 relative-luminance formula, and re-checked whenever a token changes; last verified August 17, 2026. Body text meets WCAG 2.1 AA contrast at 16.3:1. Metadata and captions meet it at 5.6:1. Inline links use a darker ochre that meets AA for body text; the lighter ochre is reserved for display and small-caps sizes, where it passes at large-text thresholds. In the footer, where type sits on ink, those roles reverse and the tokens reverse with them.

The year histogram on the index abbreviates its labels on small screens for space, but the full year stays in the page for screen readers.

A limitation of the record itself

Alt text is preserved exactly as the original site published it. Counted across the whole archive rather than a sample: of 676 images inside recovered posts, 160 carry no alt attribute at all (24%), 391 are marked decorative with an empty alt, and 125 carry description their authors wrote. Every one of the 2,193 images the archive itself emits has alt text.

Writing descriptions for the 160 would mean adding our words to the historical record, which this archive does not do. The gap is the original’s, and we preserve it rather than paper over it.

Structure, measured

Every published post — all 2,193 of them, not a sample — was checked for heading hierarchy and list markup. 2,168 (99%) passed. The exceptions are 24 posts whose author skipped a heading level and 1 carrying a list item outside any list. Both are the record’s own markup and are left as published; landmark structure and the single first-level heading come from the template, so they are correct on every page by construction.

What is still imperfect

Recovered posts carry the markup their authors wrote, across four content management systems and nine years. Some of it is untidy in ways we will not correct, because correcting it would alter the record. Where that markup makes a page harder to use, the fault is ours to route around, not the author’s to answer for.

If something here is unusable for you, please tell us what broke. We will fix what we can and say so plainly when we cannot.