1413198 - Closing a data channel directly after sending discards pending data

Status: NEW

(Core :: WebRTC: Networking, defect, P3)

Description

[Lennart Grahl [:lgrahl]]

Attachment: dc-test-send-close.html

https://tools.ietf.org/html/draft-ietf-rtcweb-data-channel-13#section-6.7 states that "[RFC6525] also guarantees that all the messages are delivered (or abandoned) before the stream is reset." This doesn't seem to be the case for Firefox.

You can test the behaviour in the attachment I've uploaded or here: https://lgrahl.de/examples/dc/dc-test-send-close.html

Comment 1

[Andreas Pehrson [:pehrsons]]

Discarding data sounds bad. Nils, could you make a more refined triage? Low P2 for now.

Comment 2

[Nils Ohlmeier [:drno]]

As Randell pointed out in #media I think this first of all needs clarification in the W3C specs to describe what should happen with the data.

Although I think that we all agree on what should be done here (attempt to send the remaining data, before closing), I would like to see at least spec issue where people appear to agree on that solution, before we start implementing a solution. Lowering priority until we have that clarification.

Comment 3

[Lennart Grahl [:lgrahl]]

I've opened an issue for the closing procedure: https://github.com/w3c/webrtc-pc/issues/1699

Comment 4

[Lennart Grahl [:lgrahl]]

It also doesn't look like the channel is being closed for the remote peer...

Comment 5

[Lennart Grahl [:lgrahl]]

This has been resolved in the spec now. I think we can proceed. :)

Comment 6

[Nils Ohlmeier [:drno]]

Right now I don't think this high enough for one of our engineers to work on any time soon.
But if you are volunteering to fix this that would be awesome! :-)