657,607 old links expose the web’s disappearing memory
A recovered database from a small 2009 link project became a 657,607-link audit of what remains reachable.
The project followed 657,607 historical links after recovering a 0.mk backup. The count turns link rot from an anecdote into a record of a web that quietly vanished.
HN’s reaction was not nostalgia alone: readers pointed out that archives are a design problem, not a promise the original web ever made.
A backup became a census
The author of the link-rot study says an old 0.mk database backup preserved 657,607 links from a side project that ran from 2009 to 2014. The team followed the recovered corpus rather than sampling it.
The archive has a business model problem
In the HN thread, the author explained why the original project closed: no revenue covered hosting, development, spam fighting, and abuse review. That detail matters because preservation is not just a storage problem. Someone has to keep paying for the service around the files.
"Webpages dying is probably one of the biggest design flaws of the original web." — a commenter recalling lost MUD logs
The phrase ‘old web’ is already contested
Readers even disagreed on what counts as old. One placed it before the mid-2000s; the dataset begins later. That disagreement does not weaken the result. It shows how quickly a normal publishing era becomes historical material once its hosts, CMSs, and domain renewals disappear.
A URL is not a preservation plan
The practical implication is modest but sharp: important material needs an archival route while it is still live. A link is an address, not a guarantee that a future reader can find anyone home.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
McDonald’s 515 pages make loyalty data feel personal
A data-access request turned one McDonald’s loyalty account into a 515-page portrait of habits, visits, and customer segmentation.