Preservation

Link Rot and Permanent URLs: Addresses Are Part of the Collection

A reference work that cannot be cited is not doing its job. Every time an address changes, every citation pointing at it breaks — in other people's articles, in bibliographies, in search results, in the notes of readers who bookmarked it years ago. The damage is invisible to the publisher and total for everyone else.

The scale of it

Link rot is not a marginal phenomenon. Studies of scholarly and legal citation have repeatedly found that a substantial share of cited web references stop resolving within a few years, and that the proportion rises steadily with age. The consequence for reference publishing is direct: a collection that changes its URL scheme every redesign is silently destroying the citations that give it standing.

The argument was made early and has not needed revising. The W3C note Cool URIs don't change was published in 1998 and remains the clearest statement of the problem, including the observation that the usual reasons given for changing URLs are organisational convenience rather than technical necessity.

Why addresses break

Almost always for one of five reasons, and none of them is inevitable.

• The technology changed. URLs carrying a file extension or a framework's conventions get rewritten when the platform does. The address encoded an implementation detail.
• The organisation changed. A department restructure becomes a path restructure. The information did not move; the org chart did.
• The title changed. Slugs derived from headings break whenever an editor improves a heading.
• The hierarchy changed. Deep paths encoding a category tree break every time an item is recategorised — and reclassification is normal in reference work.
• The site was rebuilt. The most common and least defensible: nobody made preserving addresses a requirement, so nobody did it.

Designing addresses that can survive

The general principle is that a URL should encode identity, not classification, technology or presentation. Everything that might legitimately change over the life of the collection should be kept out of the address.

• Prefer flat to deep. A single stable segment plus an identifier survives reorganisation; a five-level category path does not.
• Do not encode format or platform. The address should not reveal what generates it, so that changing the generator changes nothing.
• Keep slugs, but do not depend on them. A readable slug alongside a stable identifier gives you legibility and permanence at once, because the identifier resolves even when the slug is stale.
• Decide about trailing slashes and case once, then enforce it with redirects. Two addresses for one page split its authority and multiply the citations you must maintain.

When you must move: the discipline

Sometimes an address genuinely has to change. The obligation is then to make the old one keep working.

1. Inventory before you migrate. Produce the complete list of live addresses first. You cannot preserve what you have not enumerated, and this is the step redesigns skip.
2. Map every old address to a specific new one. Item to item, not everything to the homepage. A mass redirect of many URLs onto one unrelated page is treated by search engines as a soft error and transfers nothing — it also badly misleads a reader following a citation.
3. Use permanent redirects and keep them forever. Redirects are part of the collection now. Deleting the map in a later cleanup re-breaks everything.
4. Verify with the actual list. Check every old address resolves to the intended target. Spot checks miss precisely the long tail that citations point at.

The outbound side

Your own references rot too. A reference work citing sources that no longer resolve degrades steadily whether or not you touch it.

Two habits help. Cite in a form that survives the target moving — a persistent identifier where one exists, rather than a publisher's current URL. And where a source matters and has no persistent identifier, ensure an archived copy exists and cite that alongside the live link. The Internet Archive makes this practical for almost any public page, and the practice of citing both the live URL and a dated snapshot is now standard in careful scholarship.

Treat the address list as an asset

The most useful reframing is simply to stop thinking of URLs as an implementation detail of the website and start thinking of them as part of the published record — as much a part of the collection as the entries themselves. A project that holds a complete, current inventory of its own addresses can migrate platforms safely and repeatedly. One that does not is one redesign away from breaking every citation it has ever earned.

This is why a page that is no longer a catalogue still lives at its original address on this site. The address was cited; the address stays.