Fieldnote
Why an RID should not encode its meaning
The case for an identifier that says nothing about its subject
Look at 451.7K4M9Q2X8D5P0R6T and you learn almost nothing: a registered prefix, a dot, and sixteen characters of randomness. It is tempting to call that a shortcoming. Wouldn't an identifier that said document, Wellzai, or California be friendlier? This note argues that the opacity is the feature — the thing that lets the identifier keep its promise.
Every attribute you might encode is an attribute that can change. Documents move between spaces. Facilities change owners. Watersheds get reclassified, and administrative boundaries shift around them. A resource may even turn out to be a different kind of thing than first assumed. If any of that were written into the identifier, the identifier would either become false or have to be replaced — and a replaced identifier is precisely the broken link that persistent identity exists to prevent.
The tradeoff, honestly stated
Opacity has real costs. People cannot guess what an RID refers to, two RIDs cannot be compared by eye, and every lookup needs resolution. RSM ID pays those costs deliberately, and then gives the meaning back somewhere it can safely change. Structured, readable context lives in the RRN, which may be succeeded by a newer name while the RID stays put. Current kind, space, jurisdiction, and stewardship live in metadata returned by resolution. Even the suffix's randomness serves the same end: it is generated from eighty random bits rather than from a title, timestamp, or content hash, so nothing about the resource's current state leaks into its permanent name.
A worked case
Document 08 imagines a watershed spanning several jurisdictions with the RID 451.6R8T2W9K4M7P0Q5X. Its description may name multiple legal jurisdictions, a hydrological unit, ecological regions, and stewards. When a boundary is redrawn, the description and spatial references change; the RID does not. Had the identifier embedded a state code, the redrawing would have forced a choice between a false identifier and a new one. Revisions follow the same pattern: a revision reference identifies an immutable state, but a mutable revision number must never be embedded in the RID.
What about the kind in an RRN?
The RRN does carry a kind, and that is consistent with the argument rather than an exception to it. The kind belongs to a name, and names can be succeeded; it is drawn from a governed registry, so it cannot drift into whatever an application finds convenient. That is why the Musewoods fieldnote Code Meets Soil is named with kind document in this site's demonstration: "fieldnote" is the name of a Musewoods instrument, not one of the twelve registered kinds, and a name like rrn:451:451labs:us-ca:musewoods:fieldnote/code-meets-soil would parse but be rejected at authoritative registration.
The broader principle is easy to state and hard to keep: let the identifier answer only "which resource?", and let every other question be answered by something designed to change. An RID that says nothing is an RID that never has to lie.
What remains open
Two adjacent questions are not settled by the identifier rules alone. Document 08 allows "a substantially distinct edition" to receive a separate RID under "documented resource granularity rules", but those rules are not yet published, so when a change makes a new referent is still a policy question. And the twelve-kind registry has a minimum profile but no published entries — no definitions, stewards, or semantic-class mappings yet. Neither gap weakens the argument here; both are reminders that meaning lives in governed records that are still being written, which is exactly where it belongs.