Persistence · the daily reading
SMPTE — the body that, in 1917, fixed the width of 35 mm film and the speed it runs at — has put its entire standards catalogue online, free. Read in deep time, that is not a press release. It is a preservation act: the keys that keep a century of moving images readable have stopped quietly corroding behind a paywall.
An archivist learns early to separate two questions that look like one. The first is whether a record has survived — whether the carrier, the paper or the film or the disk, is still physically there. The second is whether anyone can still read it — whether the knowledge needed to turn those marks back into meaning has survived alongside them. The two fail independently, and the second failure is the quieter and the worse. A perfect reel of film whose format nobody remembers is a sealed jar. A standards document is the thing that keeps the jar from sealing.5 On 17 June 2026 the society that has written those documents for the moving image for a hundred and ten years gave the entire run of them away.
The Society of Motion Picture and Television Engineers announced that its whole Standards catalogue — every published Standard, every Recommended Practice, every Engineering Guideline, every Registered Disclosure Document, and all future releases — is now free to read for anyone, no membership required.1 That is the news. SMPTE's president, Rich Welsh, framed it as a turn taken at an inflection point: “Our industry is confronting transformative shifts, from IP-based workflows to AI authenticity and content provenance … Now is the time to open the gates.”1 Underneath the language of momentum sits a flatter admission, and it is the sentence I keep returning to. SMPTE's standards director, Steve Llamb, said the point of opening access is that builders “can build from accurate specifications rather than secondhand sources.”1
Hold that phrase. Secondhand sources. It concedes that a paywalled standard does not stop people from implementing the format — it only stops them from implementing it from the original. The work reroutes; it does not cease. And where it reroutes is the whole story, because for an archivist “build from a secondhand copy rather than the original” is not a modern software problem. It is the oldest problem in the transmission of texts, and it has a known failure mode. I will come back to it.
The honest reading of the announcement is that this is not a gift so much as a correction. The move is underwritten, SMPTE says, by its Diamond-level corporate members — Amazon AWS, Apple, Blackmagic Design, CBS/Paramount, Disney, Dolby, Fox, Google, Ross Video, Sony, Telstra — and by donor pledges, with a modernization of the publishing pipeline onto GitHub and structured HTML running underneath it.1 An industry that spent decades selling its standards has decided, collectively, that a standard nobody can freely read fails at the one job a standard has. To see why that is true, and not merely generous, you have to be precise about what kind of object a standard is.
Digital preservation has a formal name for the thing a format standard supplies. In the Open Archival Information System reference model — the framework most national archives and data repositories build on, published as ISO 14721 — the bits you store are called the Data Object, and they are explicitly held to be meaningless on their own. To turn a bit sequence back into a picture, a sound, a document, you need a second body of information the model calls Representation Information: the structure and semantics that map the raw bytes onto concepts a human can use.2 A published format standard is Representation Information, written down and made external. It is, quite literally, the part of the archive that lets you read the rest of the archive.
This is why the two failure modes I opened with are genuinely separate. The carrier can rot on its own schedule: nitrate film burns, acetate goes to vinegar, magnetic tape sheds its oxide, a hard drive's bits decay. That kind of loss is famous, and it is enormous — of American silent feature films, by the Library of Congress's 2013 census, on the order of seventy percent are simply gone, the negatives and prints destroyed or decomposed.3 But notice what the survivors still need. A reel that did come through is only legible if you also know the gauge it was shot on, the rate it runs at, the shape and pitch of its perforations, the way its soundtrack is encoded. Strip that knowledge away and a pristine reel is as mute as a charred one. The carrier is the jar. The standard is the key. SMPTE has spent a century cutting keys, and it has just copied the whole ring into the open.
It is worth dwelling on how deep this catalogue runs, because the deep time is the argument. SMPE — the “T” for Television was not added until 1950 — was founded in the Oak Room of the Raleigh Hotel in Washington in July 1916, by the inventor C. Francis Jenkins and nine other signatories.4 Its very first act as a standards body, at the 1917 convention in Chicago, was to fix the dimensions of 35 mm film and a frame rate of 16 per second.4 Sit with that. The society's opening document specified the literal geometry of the strip and the clock it runs against — the two facts you need, and the only two facts you need, to thread a reel from 1917 onto a machine and have it play at the right speed. Everything since has been more keys of the same kind, accreted one document at a time across a hundred and ten years.
Two of these keys deserve a moment because they show what the catalogue actually is. SMPTE 12M timecode is a metadata layer — hours, minutes, seconds, frames — laid down on every frame of a recording, an address system for time that lets a 1975 editing room and a 2026 one agree on exactly which instant a given edit touches. It is a tiny archive of its own, riding along on top of the picture. And the SMPTE color-bar test pattern is, to my eye, the most quietly beautiful object in the whole library: a signal that carries its own decoder on its face. The bars are not the content; they are a calibration target deliberately built into the format so that a receiver, decades and continents away, can reconstruct the exact conditions under which the content was meant to be read.4 It is the thing every archivist dreams of and almost never gets — the document that tells you how to read itself.
Now to the failure mode behind Llamb's secondhand sources. For most of SMPTE's life, acquiring a standard meant paying for it. One implementer on the Hacker News thread recalled the older texture of that perfectly: “acquiring a standard involved writing to a far away address and then waiting ‘six to eight weeks’ for a paper document to show up in your mailbox,” and by the time the internet arrived, “certain expectations had become concretized.”6 A price and a delay do not stop a determined builder. They reroute the builder. Another, blunt: “I have written software which needed to support SMPTE standards, and to do so I pirated the standard.”6 A third, who built cinema software, simply bought the single sub-document he needed — standard 430.10 — as a one-off PDF.6 And where a builder will not pay and cannot pirate, there is a third road, the one that matters: you reverse-engineer the format from sample files. You take a few specimens of the thing, infer the rules, and implement your guess. As one developer who works on these very codecs put it, without free standards “developers will attempt to reverse engineer from sample files, resulting in poor interoperability and causing chaos for those implementers that actually do bother to acquire and read the spec.”6
This is the oldest problem in the transmission of texts, wearing modern clothes. When the authoritative exemplar of a work is locked away, scribes copy from whatever witnesses they can reach — and, more often, from copies of copies. Each generation introduces small errors; errors compound; branches drift apart and, worse, contaminate one another. The medieval reader ended up not with the text but with a family of divergent texts, and the textual critic's whole discipline — stemmatics, the drawing of the family tree of corruption — exists because the archetype was unavailable and everything downstream had quietly mutated. A sealed-off format standard produces exactly this, except the witnesses are software. Another commenter named the offspring without flinching: the implementations become “the mutant bastard children of the existing approaches.”6 That is a stemma. That is the manuscript problem, recompiled.
Here is the irony the paywall never sees coming. Its purpose was to protect the standard — its revenue, and with it the standard's authority. But a format is, in the end, defined by what the working implementations actually do, and once those implementations are reconstructed from samples and from each other, the de facto standard becomes the cloud of divergent guesses, not the sealed text. The authoritative document, by being locked, ceases to be authoritative; it becomes a fee that a shrinking number of people pay while everyone else codifies the drift. A standard that no one reads is not a standard. It is an invoice with delusions of governance. Freeing the catalogue is how SMPTE collapses the right-hand panel back into the left.
An honest catalogue entry has to mark its own limits, so here are the ones I can see. First, open is not the same as free, and free is not the same as endowed. The catalogue is free to read now, cross-subsidized by a handful of very large companies and by pledges that are named through the end of 2026.1 That is a funding model, not a permanent settlement. A freely readable archive is exactly as durable as the institution and the infrastructure underneath it, and corporate priorities are not bedrock. The keys are in the open today; whether they stay there is a question about an institution, not about physics.
Second, the modernization is itself a bet about persistence, and it cuts both ways. Moving the documents from locked PDF to structured HTML in version control is, for preservation, a real gain: HTML is diffable, greppable, mirrorable, and — crucially — copyable, which is the one property that lets a record outlive its publisher. Anyone can now clone the texts and hold their own copy, and a record that can be copied is a record that can survive a custodian. But several builders in the thread caught the hook: hosting the standards-development workflow on a single proprietary platform reintroduces a dependency on that platform for the process — the issues, the history, the machinery of revision — even where the documents themselves are portable.6 The thing that preserves can also decay; the apparatus that makes a record legible can become the next thing that needs preserving.
Third, SMPTE is one society in a landscape that is still mostly sealed. The keys to vast stretches of the built and digital world remain behind paywalls — the IEEE 802 networking family, the ISO C and C++ language standards (whose final drafts circulate free precisely because implementers refuse to work from the paid text), the SQL specification, the National Electrical Code, the building codes a homeowner is legally bound by but must pay to read.6 The slow-erasure logic in §iv applies to every one of them. SMPTE's reversal is worth recording not because it solves the general problem but because it is a large, century-old standards body publicly conceding that the paywall was costing it the thing it was meant to protect.
And fourth, a small, sincere irony to close on. Welsh named the inflection point as “AI authenticity and content provenance” — the industry's new anxiety about where a given image truly came from.1 It is fitting, then, that the first provenance SMPTE chose to repair was the provenance of its own specifications. You cannot honestly certify where a picture came from while standing on a format the public is not allowed to read. Before you can authenticate the record, you have to unseal the key. For a hundred and ten years SMPTE kept cutting the keys to the moving image and selling them one at a time. This week it set the whole ring down where the rest of us can copy it — which is, when you read it in deep time, the most durable thing a keeper of records can do.