The Disc That Outlived Its Readers: What Domesday Still Costs Us

The most expensive lesson in British software preservation was not a catastrophe. Nothing exploded. No building burned. The data simply sat there, perfectly intact, on a disc that no longer had a machine willing to spin it.

That is the shape of format obsolescence, and it is worth staring at properly, because it is the least dramatic and most expensive failure mode in the entire industry. Nothing breaks. Things just quietly stop being openable.

A country writes itself down

In 1986 the BBC launched the Domesday Project, a snapshot of Britain assembled for the 900th anniversary of the original survey. It cost somewhere around £2.5 million — serious money then, and serious money now. More than a million people contributed. Schools wrote up their villages, their high streets, their bus routes. Aerial photographs were painstakingly digitised. It was, by any measure, a national act of remembering.

The output was two laser discs, read by a modified Philips LV-ROM player bolted to a BBC Master. Which is where the trouble started, because the project had accidentally built something magnificent and then locked it inside a reader.

The content was fine. The medium was the problem. The hardware was proprietary, niche and dependent on a single manufacturer's willingness to keep supporting it. By the mid-1990s, the players were failing, the discs were degrading, and the whole archive was heading for a shelf in a cupboard marked "later".

The rescue, and what it actually cost

The recovery that followed was genuinely heroic and genuinely slow. The CAMiLEON project, involving teams at Leeds and Michigan, worked through emulation and forensic extraction; by the early 2000s the data was being coaxed off the discs and into the open. In 2011 the BBC launched Domesday Reloaded with The National Archives, putting a great deal of it back online for the public.

It was not free. It took skilled people, years of effort, custom hardware work and institutional patience — the kind of resource that is available roughly once per generation and should never be part of a delivery plan. And even the rescue produced a website, which is a format that will itself age, quietly, in its own time.

Notice the pattern: the data was never lost. It was inaccessible. Those are very different things, and only one of them is cheap to fix.

Meanwhile, in a council near you

Fast forward to 2026. A local authority signs a seven-year contract for a case management platform. The contract is good. The diligence was done. The demo was lovely.

Ask three questions and watch the room get uncomfortable. What happens to the data at exit? In what format does it come out? Who documents the schema after the people who designed it have left?

Nobody is being villainous here. Vendors ship proprietary formats because they're convenient, optimised and commercially sensible. Councils accept them because procurement is judged on price and delivery, not on whether a hypothetical archivist in 2076 can open the file. And the cost of that decision is deferred so far into the future that it doesn't appear on any balance sheet at all. It is, in effect, a loan taken out against future citizens, with no interest rate disclosed.

This isn't hypothetical. The National Archives has spent years banging the drum on digital continuity — the simple, unglamorous discipline that says: know what you have, know what it depends on, and act before the dependency disappears. It remains one of the least fashionable things in government technology, which is a reliable sign that it matters.

The 50-year question

Here is the test I would put to every vendor and every procurement panel, and it is a beermat-sized test.

In 2076, who opens this?

If the answer is "our successors" or "the platform team" or an optimistic shrug, get something better in writing. Export in documented open formats, not just CSV dumps that lose all structure the moment a field contains a comma. Get a schema, written down, in prose, with examples. Get escrow, but don't mistake escrow for a plan — a copy of the source you can't run is just a heavier kind of unreadable. Budget for migration as a recurring line item, every seven years, forever, like a service. Take checksums so you know whether what you have is what you had. And keep one human-readable README that a person with no context could follow.

Then print it, because the one format that has never been deprecated is paper, and the one reader that has never been discontinued is a person with the document in front of them.

The quiet, expensive lesson

What is striking about Domesday is how undramatic the failure was. There was no breach, no outage, no incident channel. There was a disc, a player, and a slow realisation over about a decade that the country had written its own memory in a language that was going out of production.

The same thing is happening right now, spread across a thousand procurement decisions, each individually reasonable and collectively a slow-motion deletion. The data survives. The meaning quietly doesn't.

Which is the least satisfying kind of legacy for a country of scribblers on small surfaces. We are very good, in Britain, at making records on the back of things. We are much less good at making sure the next person can read them. It cost us millions to learn once. The invoice is being prepared again.