Media/WebRTC/2012-06-19
From MozillaWiki
Notes- 6/19/2012, 8:00am Pacific:
1) Update on getUserMedia work (Anant, Suhas)
- Current patch does audio/video/snapshot; there are issues with it sleeping the main thread.
- Delay constraints to focus on PeerConnection
2) Update on DTLS Transport work (EKR)
- I have ICE and DTLS working in unit test. Planning to land the files in my branch tomorrow (need to do a scrub for copyright issues). Next step is to integrate with SIPCC and then deal with linkage issues on Win.
- My big blocker here is how much of the code is in my branch but has not even been landed on main alder.
- Main current failings are:
- Only global configuration of STUN servers
- Need to have better prioritization of media interfaces (currently not-sensitive to interface type)
- Only TURN relaying (and that's only been minimally tested).
- Doesn't clean up properly (plan to fix that pre-landing or shortly thereafter).
3) Update on signaling work (Ethan, Enda, EKR)
- gkmedia is a pain
- PeerConnection needs to live in content/media
- Shutdown problems blocking some tests
- JSEP stuff a little behind
- Need to integrate with ICE and Media Streams
4) Update on data channels (Randell)
- Patches all updated
- Need to send email to try to resolve differences between IETF and W3
- Expect solution to SCTP->DTLS layer in next week or so
- Randell will push to keep the current design and ease the API for JS authors
5) Update on the landing plan (Randell)
- Green try with webrtc disabled (default)
- Green try except for WIndows with webrtc enabled (update RSN)
- WIll get ted to r+ minor gyp changes
- libyuv needs to come along instead of waiting for tranche 3 (no big deal)
- All other reviews done
- Plan is to land on inbound first thing tomorrow morning (6-8am) at low activity point.
6) Discuss any current blockers for people or new items that haven't been discussed
- EKR: concern about technical debt (patches not on bugs, or on bugs but not reviewed)
- Need more streamlining
- PeerConnection API issues
- gkmedia
- function pointers or pure virtual objects
- Can we move signaling to libxul?
- match up with Google
- Use callbacks instead of Observers - agreed
- ICE candidates - separate thread? trickle... they need to be async w/ respect to SDP
- Look into having SDP generation block on ICE generation (off a thread) - enda will look and ekr will dive into ICE issues if the design looks workable.
- gkmedia
- Warm up ICE candidates for normal a/v calls at peerconnection create time
- ekr/enda/ehugg/suhas will work on signaling on an alder branch
- ekr will update architecture diagram especially for ICE
7) Please update project map. Any surprises since we last talked?
- Randell did updates
- 'Green' defined (see the map)
8) Any questions/concerns about the spec work or discussions on the list? Any fallout from the interim that we need to discuss? Any questions for the Google office hours?
- None mentioned