The Archive Is a Business Decision

For most of the history of publishing, preservation was a byproduct. Print something in quantity and copies survive whether or not anyone intends them to.

Libraries formalised the accident into an obligation, and the provision permitting them to copy for preservation exists because the law recognised that keeping things is a public function.

Digital publishing removed the accident. A page exists because a company pays for it to exist, and stops existing the moment that payment stops.

There is no equivalent of a print run scattered across a thousand shelves. There is one copy, on one company's infrastructure, and its survival is a budget decision reviewed annually.

The federal records authority covers government material and covers it reasonably well. Nothing comparable governs the commercial web, which is where most of the contemporary record now lives.

The failures are not dramatic. A publication is acquired, its new owner finds the archive expensive to migrate, and a decade of work goes offline in a decision nobody announces.

A platform closes and takes with it not only its own content but every citation that pointed to it, which is how a footnote becomes an assertion nobody can check.

The scale of this is easy to underestimate because the losses are invisible by construction. You cannot notice the absence of a page you never knew existed.

The national library runs preservation programmes of real ambition, and they are necessarily selective. Capturing everything was never possible, so somebody chooses.

Selection is the whole game and it is being made by curators guessing what a future researcher will want, which is the hardest prediction in the field.

The material that survives best is the material that looked important at the time. The material historians actually end up needing is usually the ordinary stuff nobody thought worth keeping.

The copyright framework makes this harder in a way that is rarely discussed. Preservation copying is permitted narrowly, and the narrowness was written for physical objects.

An institution that wants to keep a copy of a commercial site faces a genuine legal question about whether it may, and the safe answer for a risk-averse institution is not to.

So the incentives point the same direction from both ends. The owner has no obligation to keep it and the library has limited permission to copy it.

Meanwhile the practical reality inside publishers is that an archive generates no revenue and incurs steady cost, which makes it the first thing examined when a budget tightens.

Nobody at any company decides to destroy the record of a decade. They decide not to fund a migration, which has the same effect and requires no decision at all.

The technical fixes that get proposed are real but partial. Persistent identifiers help only if somebody keeps resolving them, and resolution is another service somebody pays for.

Distributed copies help more, because they remove the single point of failure, and they require the original owner to permit copying at a moment when their interest runs the other way.

What would actually work is treating a publication's archive the way a licence treats a broadcast frequency: as something held on terms, with a deposit obligation attached.

That would be intrusive and would be resisted, and it is the only proposal on the table that does not depend on a commercial firm's continued goodwill about material it has no reason to keep.

The alternative is what we have, which is a record of this period that will turn out to be whatever happened to be cheap to store, held by whoever happened to survive.

Future historians will not describe this as censorship, because it will not have been. They will describe it as a gap, and they will not be able to say precisely what was in it.