A capture is the Internet Archive's record of one visit to one address. Everything Yore says about the past rests on captures, so it matters what one proves and what it doesn't.
What the archive records
- The time, in UTC, to the second. Yore prints it as the archive states it.
- The status the site answered with: 200 for a page, 404 for one that isn't there, 301 for a redirect.
- A digest of what it stored: the SHA-1 of the payload, written in base32. On captures since 2025 the index also states it as
sha256:followed by the SHA-256 in hex.
Every capture has an address in the web archive, which Yore links. Anyone can open it and see the archive replay that visit.
What Yore adds
Yore fetches the raw bytes of a capture and keeps them under their SHA-256. Bytes are never overwritten: the same content is stored once, and a different content gets a different name.
Yore then computes the archive's kind of digest over those bytes and says whether it matches the one in the archive's index. A match means the bytes Yore holds are the payload the archive recorded. Where they differ, Yore says so and doesn't hide the capture.
The text
The text of a page is derived from its bytes by fixed rules: headings, paragraphs, list items, table cells and the addresses of images, in the order the page has them. Scripts, styles and hidden elements are left out. No model takes part, so the same bytes always give the same text.
Yore derives two texts. The whole text is everything a reader could see. The content leaves out the site's navigation, header and footer. An export carries both, each named by its hash.
A version, and the gaps
A version runs from the first capture that reads a certain way to the last. That window is what the captures prove. Between two captures the archive didn't look: the page may have changed and changed back, and nobody would know.
So a change is dated to the capture that first shows it. Yore also states the capture before it, which is the latest moment the old text is proven. The change happened between the two.
Removals
Yore reads a removal from the archive's record of a page, by a fixed rule:
- A 404, a 410 or a 451 is an indication. Three of them, each at least 24 hours after the one before, with no capture of the page since, confirm the removal. Until then it's shown as open.
- A redirect counts once it has been held for 72 hours, across two captures or more, on the address that last answered with the page. A redirect from
httptohttps, or from another spelling of the address, is routine and never counts. - Once the archive sees the page again, the removal is resolved.
The removal is dated to the first capture that shows it. For a page you watch, Yore's own checks follow the same three-sightings rule.
Images have their own date
The archive keeps an image apart from the page that shows it, and often captured it in another month or year. When it replays a page from 2019, it fills in the nearest image it has, which may be from 2021. A court can set a capture aside for that.
So Yore reads the archive's index for each image and takes the capture nearest the version's first day. It keeps that image under its SHA-256 and states the image's own capture date, and how many days before or after the page's that is. Yore fetches the images inside a page's content, 40 a version at most, from img elements. It doesn't fetch background images set in a stylesheet.