A morning digest featured an article about why CDs and DVDs aren't eternal. Seems banal: everyone knows discs get scratched. But one phrase caught my attention — the disc's main enemy isn't outside, but inside. Scratches, it turns out, aren't even the primary cause of optical media death: what dies is the reflective layer, hidden in the polycarbonate thickness, and polishing is as useless here as a bandage for cardiac arrest.
The topic hooked me for three reasons. First, it's counterintuitive: an entire generation grew up with the ritual of wiping discs on their t-shirts, and almost nobody suspected that the t-shirt wasn't what was killing the disc. Second, behind this story stands one of the cruelest paradoxes of the digital era — the Domesday Book of 1086, beautifully preserved for nine centuries, versus its digital version from 1986, dead within sixteen years. Third, this is a story now playing out for the second time: Warner Bros. in 2025 quietly admitted that batches of their 2006–2008 DVDs "curdle like milk" — and offered replacements that may not exist.
A compact disc's construction is deceptive. Data is pressed into polycarbonate as microscopic pits, topped with a reflective layer — almost always aluminum — and covered with the thinnest lacquer coating. The entire "smart" part of the disc sits within a few dozen nanometers of the surface most accessible to the buyer. The laser during reading passes through the plastic, reflects off the metal, and returns to the photodetector: if the reflective layer degrades, the signal is lost, the drive starts making errors, and then refuses to recognize the disc altogether.
Aluminum is a treacherous material. In air it instantly forms a thin oxide film that usually protects the metal from further corrosion — this is called passivation. But a CD's reflective layer is so thin that passivation doesn't work: oxygen, sulfur, and ions carried by moisture reach the metal. Externally this looks like tiny pinhole dots visible against light, or like a stain resembling a coffee spill. Hence the term disc rot — "disc decay": a chemical process, irreversible and self-sustaining, without any owner participation.
Each format has its Achilles heel. With DVDs, where the reflective layer is sandwiched between two polycarbonate halves, corrosion protection is better — but the weak point moves to the adhesive: delamination of the "sandwich" happened all the time, especially with dual-layer discs. With recordable CD-Rs and DVD-Rs, on top of everything sits a layer of organic dye that the laser burns — and the dye decomposes over time as naturally as paint in sunlight. Rewritable CD-RWs and DVD-RWs use a phase-change alloy that fatigues even without use. The household metaphor is simple: a factory CD is a canned good, a CD-R is an opened can, a CD-RW is a can you yourself open and close each time.
The first alarm bell rang before most people had unpacked their players. In 1988 the British factory Nimbus — the UK's first CD manufacturer — published research with an unpleasant conclusion: a significant portion of discs would self-destruct within eight to ten years. The press release was accompanied by a quote from an EMI representative that today reads like a museum exhibit of overconfidence: "If you run a knife across a disc, it won't do any good, and stubbing out a cigarette on it can actually damage it. But you can pour jam on it and nothing will happen." The jam the disc truly survived. Ten years on a shelf — not always.
The most famous case of manufacturing defect was "bronzing." The Philips and Dupont Optical plant in Blackburn from 1988 to 1993 produced audio CDs with lacquer unstable to sulfur — and traces of sulfur were in the paper of inserts and booklets. Sulfur seeped through the lacquer and reacted with the reflective layer, forming bronze-colored silver sulfide. The disc would darken from edge to center, tracks at the outer radius — where the spiral physically ends — dropped errors first. Owners of classical recordings kept aesthetically beautiful but dying discs on shelves for decades: PDO for about fifteen years — from 1991 to 2006 — ran a hotline for replacing "bronzed" copies. The lesson was straightforward: one chemical incompatibility between paper and lacquer — and a legally released pressing becomes consumable material.
The BBC Domesday Project story is the perfect parable about the entire industry. In 1986, for the 900th anniversary of William the Conqueror's original Domesday Book, the BBC assembled a grand multimedia portrait of Britain: about a million people — schoolchildren and volunteers — described and photographed their towns, the data went onto interactive LV-ROM videodiscs, readable by a BBC Micro computer. The project was funded by the European Commission, software written by Logica — by 1986 standards this was a technological masterpiece, one of the first interactive multimedia in history.
Sixteen years later the original — parchment from 1086 — lay in Kew in perfect condition. Its "eternal" digital version was unreadable: LV-ROM players had died, BBC Micro computers had become museum rarities, and even physically intact discs were trapped in incompatibility. The story, however, ended happily thanks to the CAMiLEON project: in 2002 Leeds and Michigan launched an emulator of the obsolete system, and enthusiast Adrian Pearce spent sixteen months manually deciphering data structures in a hex editor — "a huge hexadecimal crossword," in his own words. Data was extracted from discs, some images recovered from original master tapes found — where else — in the project manager's attic. The final version even surpassed the original: maps, text, and photos now displayed simultaneously in three windows, something the 1986 screen couldn't afford.
The most valuable part of this parable — a quote from project manager Paul Wheatley, said in 2002 and sounding today like a law of digital physics: "It is mistaken to assume that recording on a durable medium provides meaningful preservation." The disc can be physically intact — and dead to the world, because the ecosystem of reading has disappeared around it.
There's honest science about how discs die. Standards ISO/IEC 10995 and ECMA-379 describe accelerated aging methodology: samples are kept in chambers under harsh conditions — around 80–85 °C and 85% humidity — for hours, tracking the growth of error rates (BLER for CD, PI8 for DVD), then Eyring and Arrhenius mathematics extrapolates chemical kinetics to normal conditions to obtain "estimated lifespan." This is the same class of methods automakers use to predict polymer aging — with the same fundamental caveat: the model accounts only for temperature and humidity, not dust, sulfur in paper, stickers, or fate's malice.
The classic joint study by NIST and the Library of Congress (Special Publication 500-263, 2005) compared recordable media from different manufacturers under identical conditions and got variation more insulting than any marketing cover: some samples maintained low error levels to test end, others degraded several times faster — with outwardly similar "blanks." The Canadian Conservation Institute condensed the accumulated knowledge into a table where variation by media type ranges from "over a hundred years" to "five to ten years": leader — CD-R with phthalocyanine dye and gold reflective layer, laggards — dual-layer DVD+R DL and dye-based BD-R, living 5–10 years. An amusing detail for those who considered Blu-ray an evolution in reliability: according to Iraci research (2018) BD-Rs proved on average less stable than good CD-Rs and DVD±Rs — manufacturers optimized density, not longevity.
From this same science grew the main practical wisdom of burning: write speed matters. Discs rated for 52× give minimum error levels when written at 4–12×; rushing the write is like a smudged ink signature that's unreadable in principle after twenty years. And defect quality localizes to specific plants: the famous LaserDisc "laser rot" afflicted output from specific factories — early MCA DiscoVision pressings, Sony DADC press in Terre Haute — so the collector community kept lists of "cursed" pressings, like seismologists keep epicenter maps.
The industry tried to answer rot with a direct engineering move. M-DISC, introduced in 2009 by Millenniata, abandoned organic dye altogether: data is "engraved" by a high-power laser into an inorganic glassy layer that theoretically has nothing to oxidize. Marketing promised thousand-year storage, US DoD tests at China Lake (2009) — where all organic-dye discs failed the accelerated test series — looked convincing, and a 2016 consumer experiment where M-DISC after two months outdoors read error-free while ordinary BD-R was physically destroyed went viral as proof.
Then begins the very irony worth digging for. Millenniata went bankrupt in 2016; the brand and manufacturing flowed to CMC Magnetics and Verbatim, after which quiet paranoia unfolded in the archivist community: scanning media IDs of fresh M-DISC Blu-rays showed matches with identifiers of ordinary quality BD-Rs — suspicion arose that the "stone layer" had been replaced in places with standard inorganic alloy, selling a premium brand without premium chemistry. Plus a limitation the manufacturer itself admits: any disc's polycarbonate base is rated for about a thousand years, so even a perfect data layer will die along with the plastic. NIST IR 8387 (2022) ultimately cites M-DISC as an archival format of "100+ years" — a figure honest and boring, which marketing has already abandoned and physics hasn't yet reached.
Most tempting — to close the topic with "discs die, therefore everything to the cloud." A 2024 Pew Research Center study coldly counts the opposite: 38% of web pages that existed in 2013 were inaccessible by 2023; 23% of news pages have at least one broken link; 54% of Wikipedia pages contain at least one dead source in the references section. This is link rot — decay not of the carrier but of the address, and it happens even faster than chemistry in plastic. Digital decay isn't canceled by cloud migration — it just changes carrier: now the provider can rot.
An illustration at the "can't believe it's 2026" level: St. Louis TV station Nine PBS entrusted 50 TB — seventy years of television history — to Open Source Storage cloud storage, whose infrastructure stood in Iron Mountain's Denver data center. In February 2026 the provider stopped responding, in March — cut off access; it emerged the company was listed as delinquent with Colorado's Secretary of State, and someone named Tramel who'd bought OSS then claimed he'd been "deceived" in the purchase. Hardware belonged to the provider, data center to a third party, data to the station: court ultimately confirmed ownership and ordered Iron Mountain to cooperate, but restoring access took months and two lawsuits. Meanwhile the industry quietly returned to tape: LTO cassettes, which the mocking intelligentsia had written off for ten years, shipped 160.3 exabytes of capacity in 2025, and the LTO Consortium forecast is new growth by end of 2026, because AI infrastructure needs cheap cold storage, not promises of eternity.
My main discovery in this topic — not chemistry, but semantics. All talk of carrier reliability reduces to the habit of measuring object lifespan, whereas what's always preserved is a system: carrier + reading device + format + software + institution that remembers and reproduces all this. CDs die by the first letter of this chain, Domesday died by the remaining four, Nine PBS nearly died by the last. Longevity isn't a property of material but of maintenance: "ink on paper" lives for centuries not because paper is nobler than polycarbonate, but because reading paper doesn't require an entire industry to remain compatible with itself.
Second conclusion — about the false dichotomy "physical vs. cloud." Discs rot chemically, the cloud rots legally and economically, and choosing "eternal" storage is impossible in principle — you can only buy time until the next migration. The rule of three copies on two media types with one copy off-site, which archivists repeat like a mantra, isn't paranoia but honest acknowledgment that each technology has its own failure class. Amusing that humanity's most durable message — Voyager's golden record — solves the task the same way: physically inert carrier and an entire team that wrote reading instructions — it's just this team is designed for five billion years of operation without staff replacement.
And final irony for nostalgia lovers: by all conservation tables the most durable recordable disc is the modest late-nineties CD-R with phthalocyanine and gold coating, product of a long "dead" format. Progress in data storage — like in clothing: newest doesn't mean best, sometimes it's just a different compromise between density, price, and shelf life.