The stateful GET stream and the stateless subscriptions/listen stream
capture the client address when they open and resolve every periodic
re-check with towel.WithClientIP, as the bearer middleware does, so a
failing re-check (a revoked admin token, for example) is logged with the
remote address. TestStreamRecheckClientIP records the address the verifier
sees on every verification of both streams.
Each stateless request now carries an entry-use record on its context;
getServer counts the entry it hands out as active (under the path lock)
and ServeHTTP releases it when the request ends, so a subscriptions/listen
stream keeps its entry active for its whole life. Capacity eviction walks
the LRU from the back and drops only idle entries, so it no longer ends a
quiet listener or cancels a tool call in flight, and a revocation still
finds every open stream. While every entry is busy the LRU stays above
capacity until requests end.
TestStatelessFanOut's LRU subtest now asserts that the listener's server is
spared when the binding past capacity arrives (the oldest idle server goes
instead) and that a revocation then still ends the stream.
catalog.allowedReport returns the allowed lists together with the first
provider error. Widening (stateful) and the stateless refresh leave the
cached server and its recorded fingerprint untouched on an error, so the
next request retries instead of the tools staying hidden until the next
rotation or event. A fan-out recompute still applies its fail-closed diff
but records no fingerprint, so the binding's next request re-filters.
TestWideningGuardError fails the guard during a widening on both paths and
during a permissions recompute, then asserts that the next request lists
every tool once the provider answers again.
The identity fingerprint no longer includes token ids: guards decide on
scopes, subjects and the ceiling, and an OAuth access token's id changes on
every refresh, so a rotation alone no longer re-filters a cached server.
The stateless path's refresh only adds tools and prompts for a fingerprint
new to the entry (remembered until a fan-out recompute) and never removes:
one server serves every live token of a binding, and removing on each
alternation between a wider and a narrower token pulled a tool from under
the wider token's call. The receiving middleware still filters every list
and call per request; fan-out recomputes still narrow. The stateful path
keeps its exact widen, a session serves one client.
TestFingerprintIgnoresTokenID pins the fingerprint; TestStatelessSharedBinding
blocks the wide token's call after its server was chosen, sends a request
of the narrow token and asserts the call still succeeds (it failed with
"unknown tool" before).
Fan-out matching snapshots the sessions (and stateless entries) under the
router-wide lock, releases it, and reads each identity through a small
identity box that is never held across I/O, so an event no longer waits on
a session that is re-filtering while it holds sr.mu or sp.mu.
Widening and the fan-out recomputes run their guard checks with
guardContext: not cancelled with the triggering request, but bounded by
mcp.call_timeout, so a provider that never answers cannot hold a session
lock forever. README and docs/services/mcp.md describe the wider meaning of
mcp.call_timeout.
TestLockScopeUnderWidening blocks a widening guard check, fires a
permission event for that session and asserts that another session's call
still finishes, and that a never-answering check ends at call_timeout.
- pact.MCPSubjectChanged and pier.Event gain TokenID and ClientID; a
narrowed revoked event closes only matching sessions and stateless servers
- cabana fires the admin token revoke with the token row id
- towel.WithClientIP/ClientIP; pier puts the address (surf applies
http.trusted_proxies) on the resolve context and ResolveBackendToken logs it
- Identity.fingerprint covers scopes, both subjects, the ceiling and the
token kind; a stateful lookup or stateless hit whose fingerprint changed
re-filters the server with the event diff under its own mutex (A-WR-04)
- Mount warms a Redis store in the background with a 1 s to 30 s backoff
and every request kicks a pending open without waiting, so stateless-only
instances subscribe to fan-out (A-WR-05)
- a stateless subscriptions/listen POST re-verifies its bearer every
recheck interval; an LRU-evicted stateless server closes its sessions
(A-WR-06)
- resolvePersonalToken refuses a subject with TokenID 0, and the
pact.MCPPersonalTokenResolver contract says one resolver per application,
false means unauthorized and TokenID keys the binding (A-WR-09, A-IN-03)
- an unexported reserver (Reserve, Release) next to updater: the memory
store reserves under its mutex, the Redis store in one Lua script
(prune, SCARD, SET PX plus SADD); serveNew reserves instead of Count
then Put and releases on a failed connect or initialize (A-WR-03)
- serveNew, lookup's rehydration and the stateless getServer build the
SDK server with no router-wide lock held and insert or adopt under the
lock afterwards, so a slow guard or database check no longer stalls
the instance (A-WR-07)
- store_redis.go: records with PX TTLs, per-binding sets, pub/sub events,
opened lazily from redis.url under cooler's prefix (nothing dials at mount)
- events.go: pact.MCPSubjectChanged republished through the store; every
instance recomputes the subject's sessions and stateless servers
(list_changed) or ends them on revocation; the link cache is dropped
- stateless.go: the SDK's stateless handler over per-identity servers for
2026-07-28 clients, routed by the protocol header or server/discover
- prm.go: RFC 9728 metadata at both well-known paths; long GET streams
re-verify their bearer every minute
- TestPierSessionsSmoke, TestPierFanOutSmoke, TestPierStatelessSmoke on
miniredis; README features, configuration and operational notes
