Closed guoye-zhang closed 3 years ago
I would appreciate some quick discussion if this issue is important to address before I approve or not.
We could make this a MAY. The other formulation we've used is "MUST ... if the client detects", making the response mandatory if the condition is detected, but not specifying the required level of effort to detect it.
I don't know that this issue is critical to address, but I agree that with certain (reasonable) implementation strategies this particular MUST becomes unenforceable / unachievable.
+1 for "MUST if the client detects" if we're going to make a change.
I think "MUST if detects" is suitable here - it points at the core problem. Simply s/MUST/MAY doesn't help people appreciate that might not be able to detect the condition. I'd also suggest that "MUST if detects" is an editorial change.
@guoye-zhang can you confirm if #4829 addresses your issue?
Looks good!
Thanks. In that case I believe we have addressed this issue with an editorial change. @gloinul please confirm if we can merge the related PR and incorporate the change in during the RFC editor process.
Yes, the change looks good. I could actually include it as an RFC-edtior note in my approval message.
For clients that have sent CANCEL_PUSH, they don't know if server's push stream will still arrive or not. They have to remember Push IDs forever in order to differentiate between cancelled push and completed push.