Somebody assembles a list of references, posts it, and it gets copied everywhere. Two years later most of it does not work. This happens to every such list, without exception, and it happens through a small number of identifiable mechanisms. Understanding them tells you which lists are worth keeping and how to build one that ages better than the rest.

The four ways a list breaks

Deletion is the steady background rate. Items come down for takedown requests, duplicate cleanup, policy changes, or an uploader removing their own post, and each one silently punches a hole in every list that mentioned it. Nobody is notified, so the list looks fine until somebody tries an entry.

Migration is rarer and far more destructive. A library rebuilding its storage may keep identifiers, remap them, or abandon them entirely, and only the first option leaves existing references intact. That decision is made by engineers who are not thinking about a decade of forum posts.

The first is gradual and the second and third are sudden. The fourth is the one people cause themselves, and it is the reason a broken list can circulate for years after everyone who could have fixed it has moved on.

Copying is what really kills them

The original author might maintain their list. Nobody who copies it will. Each copy is a snapshot taken at one moment, republished without dates, and treated as current by everyone who finds it afterwards. Within a few generations of copying, there is no way to tell how old any entry is.

This is why identical broken lists appear on dozens of sites. They are not independent sources agreeing with each other; they are one source photocopied, which is the same false-consensus problem described on the clone-site family page.

Adding a date to a list solves most of this, and almost no list has one.

Bare numbers cannot be repaired

Here is the practical difference. A list entry reading a number and nothing else is unrecoverable once it stops resolving — there is no information left to search with. An entry with a title and artist can be found again on any other library, in any translation, for as long as the work exists anywhere.

That single formatting decision determines whether a list has a lifespan of two years or twenty. It costs one extra column, and it is the argument running through codes, tags, and titles.

Judging a list you find

Check three things. Does it carry a date. Does it include titles as well as references. Does anything indicate it has been revisited since it was written. A list failing all three is a historical document rather than a resource, however useful it looks.

Provenance helps too. A list on a personal page by someone who still posts is more likely to be maintained than the same list reposted on an aggregator with no author attached. Knowing who made something is most of knowing whether anyone is looking after it.

Spot-check five entries before trusting the rest. If two are dead, the list is old enough that its recommendations reflect a different moment in the medium as well as a different set of live references.

Building one that ages better

Date it. Include titles and artists. Say when you last checked. Keep it short enough that rechecking is realistic, because a list of thirty entries can be verified in an evening and a list of six hundred never will be.

If you publish it, say plainly that it is a snapshot. That one sentence prevents the copying problem from compounding, because the next person knows to check rather than assume.

What a list of references is worth

Quite a lot on the day it is written and steadily less afterwards, unless it was built to survive. The difference is entirely in whether it recorded what the works are as well as where they were, and that is a decision made in about five seconds per entry.

The personal version of the same practice is on building an index you actually keep, and the standard for what belongs on any such list is on the understanding of this house.