The Century That Forgot Itself

In 1086, a survey of England was compiled and written out by hand on sheepskin. It recorded who held which land, how many ploughs and mills and swine, and what it had all been worth twenty years before. It was bound, stored, and consulted for centuries. It survives. You may, with appropriate arrangements, go and look at it, and read it, without any equipment whatsoever beyond a knowledge of medieval Latin and reasonable eyesight.

In 1986, to mark the survey's nine-hundredth anniversary, a vastly more ambitious project was undertaken. A million people contributed photographs, maps, statistics and written accounts of their own localities. The result was pressed onto twelve-inch optical discs, read by a specific microcomputer connected to a specific disc player through a specific interface, and it was magnificent — interactive, searchable, richer by every measure than the parchment it commemorated.

Within sixteen years it was effectively unreadable. The computers had gone. The drives had gone. The people who understood the file structures had moved on or died. Recovering it required a specialist rescue effort and several years of work, undertaken by people who understood, with some urgency, that they were the last cohort who could.

The parchment had lasted nine hundred years. The discs had lasted less than twenty. This is not a paradox. It is the ordinary arithmetic of the medium, and we have not begun to reckon with it honestly.

---

We tell ourselves a story about digital information, and the story is that it is permanent. Copies are perfect. Nothing fades, nothing yellows, nothing is eaten by insects in a damp attic. A file duplicated a thousand times is a thousand identical files, and the internet — that vast, humming, redundant thing — remembers everything, sometimes more than we would wish.

Almost every part of this is wrong, and the error is instructive, because it mistakes a property of the representation for a property of the artefact.

It is true that a sequence of numbers can be copied without loss. But a sequence of numbers is not a document. To become a document it requires a physical substrate that holds the numbers, a device that can read that substrate, an interface by which the device connects to something, a piece of software that interprets the resulting bytes as a particular format, and an operating system on which that software will run. Six dependencies, each with its own lifespan, and the document survives only as long as the shortest of them.

Consider each briefly and the position becomes uncomfortable. Magnetic tape loses its coating and its readability over decades, and tape drives typically read only one or two generations back, so an archive must be recopied to new media on a rolling schedule forever — an unbroken chain of institutional attention that no civilisation has ever sustained for anything. Writable optical discs, once sold as archival, rely on organic dyes that degrade under heat and light, sometimes within a decade. Flash memory, left unpowered, leaks charge and forgets. Hard drives fail in ways that are well characterised statistically and entirely unpredictable individually.

And the formats. A format is a convention — an agreement about what the numbers mean — and conventions are maintained by institutions with commercial interests that change. Word processors that dominated their era and produced millions of documents are gone, and their files, though perfectly preserved as bits, are now meaningless to nearly everything that might open them. A multimedia runtime that carried an entire creative decade was discontinued on a scheduled date, and a body of work approaching the size of a national collection became, overnight, playable only under emulation.

None of this is decay in the familiar sense. The bits do not rot. What rots is the context that made them legible — and context, unlike parchment, leaves no trace of its own disappearance. A corrupted photograph looks corrupted. An unreadable file looks like nothing at all.

---

The losses we are actually sustaining are more banal than the dramatic framing suggests, and worse.

Studies of scholarly and legal citation have repeatedly found that a large fraction of referenced web addresses no longer resolve, and that a further fraction resolve to material that has silently changed since it was cited. The reference survives; the referent does not. An argument built on such a citation cannot be checked, which means it has quietly stopped being an argument and become an assertion.

Personal memory has undergone the same transformation without anyone deciding it should. The photographs of the twentieth century survive in shoeboxes, unsorted, badly stored, and entirely readable — you open the lid and there they are. The photographs of the last twenty-five years are on a laptop that failed, a phone that was replaced, a service that closed with ninety days' notice, or an account whose password died with its owner. We have taken more images in a single recent year than in the entire history of chemical photography, and we have made almost no arrangements for any of them.

---

What can be done is known, and it is neither mysterious nor especially expensive. It is merely relentless.

Keep many copies, in different places, on different media, under different administrations, and check them against each other continuously so that a failure is discovered while the other copies remain. Prefer formats that are openly specified, so that a reader can be written from documentation if every existing reader vanishes. Migrate deliberately, on a schedule, rather than reactively when something breaks. Maintain, alongside the data, a description of what the data is — because an undocumented archive is a warehouse of unlabelled boxes, and the labour of opening them will eventually exceed anyone's willingness to pay for it.

All of this is ordinary institutional maintenance. It requires no innovation. It requires only that someone keep doing it after the project that created the material has ended, the funding has moved, and the people who cared have retired — which is precisely the thing our institutions are worst at.

---

There is a particular irony in being the best-documented generation that has ever lived and simultaneously the most likely to vanish. We produce more record per person per day than a medieval kingdom produced per year, and we store it on substrates with the durability of fruit.

The monk copying a manuscript understood something we have allowed ourselves to forget: that transmission is not a property of the object but an activity performed by the living, repeatedly, in every generation, for as long as anyone wants the thing to last. He knew the vellum would outlive him and that this was not sufficient. Someone would have to copy it again.

We have built the most capacious memory in human history and forgotten to appoint anyone to keep it.