Re: [quicwg/base-drafts] Clearer text for application errors (#3226)

[Mike Bishop <[e-mail]>]

MikeBishop commented on this pull request.



> @@ -2720,17 +2720,19 @@ connection in a recoverable state, the endpoint can send a RESET_STREAM frame
 ({{frame-reset-stream}}) with an appropriate error code to terminate just the
 affected stream.
 
-RESET_STREAM MUST be instigated by the protocol using QUIC.  RESET_STREAM
-carries an application error code.  Only the application protocol is able to
+RESET_STREAM MUST be instigated by the application protocol that uses QUIC.

```suggestion
RESET_STREAM MUST only be instigated by the application protocol that uses QUIC.
```
@mikkelfj is right -- in isolation, this reads like a requirement on the application, not a restriction on everyone else.

>  cause a stream to be terminated.  A local instance of the application protocol
 uses a direct API call and a remote instance uses the STOP_SENDING frame, which
 triggers an automatic RESET_STREAM.
 
-Resetting a stream without knowledge of the application protocol could cause the
-protocol to enter an unrecoverable state.  Application protocols might require
-certain streams to be reliably delivered in order to guarantee consistent state
-between endpoints.  Application protocols SHOULD define rules for handling
-streams that are prematurely cancelled by either endpoint.
+Resetting a stream without the involvement of the application protocol could
+cause the application protocol to enter an unrecoverable state.  Application
+protocols might require certain streams to be reliably delivered in order to

It's the "critical stream" idea from HTTP/3 -- if you blow up a request stream, the HTTP connection is fine.  If you blow up a QPACK stream, the connection dies in flames.  Some streams are application-critical, and only the application knows which, so only the application can decide which streams to kill.  In a transport world, though, "reliably delivered" brings up different overtones.  Maybe "Resetting certain streams could disrupt the consistency of application-layer state between the endpoints."?

As to confusion, this is the rationale for the restriction in the previous paragraph.  Maybe putting the rationale first and following with "Therefore, RESET_STREAM MUST only be instigated...." would help?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/pull/3226#discussion_r345360528