Soupspoon wrote:That best practice doesn't rely on continual maintenance. It more works towards spreading the bets so that any unforseen pause in the process will still leave various mirroring versions of archival material around (and probably around in different places, each differently vulnerable to the vicisitudes of time) so that once the future turns back to an interest in the old material there's a multi-pronged possibility of finding something recoverable. Either directly or by piecing together fragments from across the canon. Moreover, it is not inconceivable that fragments of acetate-stored records, together with a hardy but lost-art digital representations and various other breadcrumbs could form a Rosetta Stone towards the better understanding of a yet wider range of materials, given the opportunity for further study.
That's ignoring the more immediate recovery from "Oh no! All our CD-RW archive media from 15 years ago have degraded!" problems, necessitating pulls from the bulk deep-paper archives and/or Last Good Read copies to more contemporanius (or at least more recently re-written) forms of media.
These discussions often confuse two different things, though. There's the longevity of the medium, which is what people are referring to when they say "oh, yeah, they told us CDs would last forever". We've obviously realised that individual digital media can't be trusted, and that we need to keep backups in multiple places and continually migrate them onto the latest technology, as you describe. But the great thing about digital files is that this is extremely easy and cheap to do, because it doesn't matter what the data is: providing reliable storage for digital files is a standard off-the-shelf service. It doesn't need specialists in particular video tape standards or microfiche formats. The files themselves are the archives, not the magnetic discs or whatever on which they're stored.
That leaves the problem of longevity of the file format, which I think is what's being discussed here and is the major problem. Nevertheless, I'd expect that text
should last as long as the writing systems themselves survive, provided there's a way of pulling the wheaty text out of all the other chaff in the file. (Diagrams and photos are another story). This implies that a good format ought definitely to preserve the order of text characters and keep passages of text together. PDFs don't always do this; quite often the letters of a word will be separated from one another (I know this from trying to search a PDF for a word that I can damn well see occurs at least once).