Why do old files become unreadable, and what is digital preservation?
Because a file is only meaningful to software that understands its format, and both the software and the medium it sits on decay — on timescales far shorter than people assume.
The three separate failure modes, which get conflated:
Bit rot — the medium fails. Magnetic media lose charge, optical discs delaminate, flash memory cells leak charge when left unpowered for years, and mechanical drives seize. Consumer storage is not archival storage, and the manufacturer's claimed lifespan assumes conditions nobody actually maintains.
Format obsolescence — nothing can open it. The bits survive perfectly and no current software understands them. Proprietary formats from defunct companies are the classic case: word processor documents, CAD files, early image formats, and anything tied to a plug-in.
Dependency loss — it opens but is wrong. The file references fonts, linked files, external databases or a specific software version. It renders, incorrectly, and the damage is subtle enough to go unnoticed.
Why cloud storage is not a solution by itself. It addresses medium failure well, through replication, and addresses format obsolescence not at all. It also introduces a new dependency: the service continuing to exist and your account continuing to be accessible.
What actually works:
Open, documented formats. Plain text, PDF/A, TIFF, PNG, FLAC, CSV. The specification being public means a reader can be written again later even if every current implementation disappears.
Multiple copies in multiple places on multiple media types — the standard archival principle, because failure modes are rarely simultaneous.
Periodic migration. Not storing forever but moving forward — checking and converting as formats age. This is the part people skip, and it is the part that matters.
Checksums recorded alongside files, so silent corruption is detectable rather than discovered when you open the file.
Documentation of context — what the file is, what made it, what it needs.
Institutional archives exist for precisely this problem, and emulation is increasingly used to run obsolete software rather than convert its output.