Re: [quicwg/base-drafts] Acknowledge stream resets too (#1400)

[Mike Bishop <[e-mail]>]

I think we may need to specify this a little more precisely, either here or in a separate section.

There are two scenarios to care about:

- **Stream has partially-received data, then gets a RST_STREAM.**  Transport+decoder might abort any unread data and immediately surface the RST, or it might save the RST for the first gap in received data, if any.  If the decoder is reading the data, perhaps it SHOULD acknowledge any blocks that it processes, then MUST send the RSA when it will not consume any more header blocks.

- **Receiver decides to send a STOP_SENDING.**  HTTP implementation/decoder decide that no more data will be consumed from a stream for some reason.  Even without receiving the RST_STREAM the STOP_SENDING will trigger, the decoder should emit this at the point no more data will be consumed from the stream.

Maybe we should phrase it as being sent when no further data on a stream will be consumed, but the stream was not read to completion.  Then receipt of a RST_STREAM and issuing a STOP_SENDING are cases that will happen.

-- 
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/1400#issuecomment-393248033