Identifiers

Every article, principal and revision has exactly one identifier:

01K3XJ9V2QF8H0M4T6RZC7NB5D

26 characters, Crockford base32, from a UUIDv7. Three properties follow from that, and all three are load-bearing.

It sorts by time

UUIDv7 puts the timestamp first, so lexicographic order is creation order. That is why pagination on this API is a cursor holding the last id you saw, and never an offset — see Consistency.

It never changes and is never reused

Not on rename, not on republication, not after deletion. A deleted article’s id keeps its place in the citation graph and answers 410 Gone rather than 404, so a link written before the takedown still says something true — this existed, and is not here any more — instead of quietly becoming a lie.

There is only one of it

No internal id with a public alias beside it. A pair means two things to keep in step, two places to leak the wrong one, and an eventual migration when somebody discovers the internal one in a URL.

The URL contains nothing else

https://orator.space/p/01K3XJ9V2QF8H0M4T6RZC7NB5D

No slug. A slug is derived from a title, titles get corrected, and a link is a promise — so either the slug goes stale or the correction is refused. Anything appended to a valid id redirects back to the canonical form, so a link written before that was true still resolves.

The same id, three representations:

/p/{id}         HTML for a person
/p/{id}.md      the markdown, as the author sent it
/p/{id}.json    the article and its metadata

Separate URLs rather than content negotiation on one, so that a cache, a link and a citation all refer to exactly one thing.