Finish the realm→domain terminology removal across source, tests, e2e
and docs. The stored config drops all lists: Config.domains is keyed by
rp-id, DomainConfig.origins/related are objects keyed by host (https://
omitted), values True or OriginEntry(auth_host=True). The default/primary
domain concept is gone; ordering is display-time. Tests and e2e updated
to the new API shapes (not run). Database re-migrated from the legacy
backup into the new format.
Serving never refuses to start because of stored realm config: the
registry build sanitizes best-effort and warns — misfiled origin entries
are reclassified (a cross-domain origins entry is served as a related
origin) or dropped, collisions resolve first-come-wins, over-cap related
lists truncate, unsalvageable realms are skipped. Fixing the stored
config stays the admin interface's job, and it stays reachable on any
working realm. Only a config with no servable realm at all is fatal.
Admin realm writes stay strict and gain self-lockout guards: an update
that would leave the admin's current host unable to run ceremonies for
the realm they are on is refused (unless an auth host takes over
ceremonies), and deleting the realm currently in use is refused.
RealmConfig.origins is again purely an allow-list of sign-in sites
within the realm's domain (unset = rp-id and all subdomains), restoring
the restriction semantics the realm rework had silently turned into an
always-open subtree. Cross-domain ROR origins move to their own
RealmConfig.related_origins field — always additive, capped, validated
to be outside the rp-id domain, and the sole source of the
/.well-known/webauthn document.
Admin API POST/PATCH accept related_origins; misfiled entries are
rejected (cross-domain in origins, in-domain in related_origins).
Admin UI: the realm dialog edits the two lists separately with
end-user-oriented explanations (allowed sign-in sites vs. related
domains + the well-known note); the Realms section intro explains the
multi-domain model, and the table shows sign-in site and related domain
counts.
Serve never converts databases: with no paskia.kantadb it points at
'paskia init', or at 'paskia migrate' when legacy *.paskiadb candidates
exist. migrate converts a lone candidate, or the one named by --rp-id
when several exist; the rest stay in place. devserver fails fast with
the same hint. gitignore covers paskia.kantadb and *.converted-bak.
README quick start and configuration updated for init/serve split and
paskia.kantadb; API.md server-config endpoints replaced with admin realm
CRUD and the well-known webauthn endpoint; proxy guides point at the
realm auth-host setting; oidc.md documents per-realm issuers; MultiSite.md
rewritten from the implementation plan into documentation of the shipped
mechanics, policy model, and design rationale.