The digital dilemma

In 2006 the Academy of Motion Picture Arts and Sciences asked its Science and Technology Council to study the state of digital preservation across the industry. Milt Shefter, a preservation specialist, led the resulting Digital Motion Picture Archival Project and wrote and edited the report with Andy Maltz, the Council's director. The Digital Dilemma came out in November 2007 as the Academy's first white paper. It won the Society of American Archivists' Preservation Publication Award the following year, went out in more than six thousand copies, and was translated into Japanese, Brazilian Portuguese and Korean. The number people remember is the cost comparison. Keeping a film archival master, the report found, ran to about 1,059 dollars a year per title. Keeping a 4K digital master of the same title ran to about 12,514 dollars a year, roughly eleven times as much, and that figure assumed fully managed storage of three copies, not a drive on a shelf. The reason for the gap is not that hard drives are expensive. It is that film sits in a cold room and asks nothing of anyone for decades, while data has to be copied onto new hardware every few years, checked for errors, and moved to new formats before the old ones stop being readable. The report's phrase for it was that digital data, unlike film, cannot survive long periods of benign neglect. Its authors wrote that the issues of storage reliability and compatibility go far beyond lost snapshots and spreadsheets, which was a polite way of saying that the industry had no plan. The Council followed up in 2012 with The Digital Dilemma 2, produced with the Library of Congress's digital preservation program and aimed at the people the first report had not reached: independent filmmakers, documentarians and nonprofit archives. It found that they lacked the resources, personnel and funding to keep their own work, at a time when three quarters of theatrically released films were independently produced.

Film survives neglect. Digital survives only attention.

A household with a box of tapes and a folder called CAMCORDER is a very small independent archive, and it has exactly the problem the second report describes. Nobody is paid to look after it. The files exist on one drive, the drive is in a drawer, and the cable that connected the old camera to a computer fits nothing anyone owns now. This is worth being blunt about. A digital copy is not an archive. It is a copy, and it lasts precisely as long as someone keeps copying it forward. The scan you paid for is the beginning of the work, not the end.

The rule is short and people misquote it. It says: keep at least three copies of the data, on two different types of storage media, with one copy kept somewhere else. Three copies on three drives from the same shop in the same drawer is one copy with extra steps, because they fail the same way and burn in the same fire. A household version that meets the rule looks like this. Copy one is the working folder on the computer. Copy two is an external drive, a different medium, kept in the closet with the reels. Copy three is either a cloud account or a second drive at a sibling's house, which is the offsite part. That is the whole structure. The routine that keeps it alive is small. Twice a year, plug in the drive and open a handful of files from different years. Every few years, buy a new drive and copy everything across, then keep the old one until the new one has been checked. When your current computer starts to struggle with the files, or the format needs a special program to open, convert them then and keep the originals beside the conversions. Write the date of each check on a card in the box. The card matters more than it sounds, because the person who eventually inherits the drive will otherwise have no idea whether anyone ever looked.

A checksum is a fingerprint for a file. A program reads every byte and boils them down to a short string of letters and numbers. Run it again on the same file and you get the same string. Change a single bit anywhere in the file and the string comes out different. Archivists call the practice fixity checking, and it is the only way to know that a copy is still the file you made rather than one that looks the same and plays with a glitch at minute nine. Free tools do it, on every operating system. You make the list of fingerprints once, keep the list next to the files, and compare when you do the twice-yearly check.

Bit rot is the reason to bother. Data degradation, to give it the technical name, is the gradual corruption of stored data through an accumulation of small failures that are not individually fatal. On a hard drive, bits slowly lose their magnetic orientation and heat speeds the loss. In flash memory the stored charge leaks away through insulation that is never perfect. On optical discs the medium itself breaks down. It is real, and it is slow: current drives are rated with magnetic lifespans measured in decades, so a file is far more likely to be lost to a dead drive or a lost login than to rot in place. The catch is that rot is silent. A drive does not tell you that one frame in a file it has held for eight years now has a wrong value. The published position is that no method eliminates the risk, and that the working defense is checksums to detect it plus multiple copies in more than one place to repair it. That is what the rule above gives you. Three copies, checked on a schedule, with fingerprints to compare, is a plan an archive would recognize. It costs a couple of drives and an hour twice a year.

A digital copy lasts exactly as long as someone keeps copying it forward.