Minutes for RTCWEB at interim-2016-rtcweb-1
minutes-interim-2016-rtcweb-1-1

RTCWEB Interim meeting
March 10, 2016
Chairs: Cullen Jennings, Ted Hardie (regrets from Sean Turner)

Attendees:
Cullen Jennings
Dan Romascanu
Justin Uberti
Taylor Brandsetter
Peter
Bernard Aboba
Alex Gouaillard
Magnus Westerlund
Harald Alvestrand
Ben Campbell
Adam Roach
Alan Johnston
Randell Jesup
Maire Reavy
Roman Shpount
Victor Pascual
Ted Hardie

Agenda:

Review draft-ietf-rtcweb-jsep-13 and discuss changes (see changelog in Appendix
A and github issues list).

Action item: land PRs 215 and 219, emit new draft,  and send message to the
working highlighting changes.  The transceiver issues are the largest pole left
in the tent.

Prior to starting the main JSEP discussion, the group clarified which SDP lines
can be ignored:  effort to identify which SDP lines have no meaning in this
context (e.g. time of session start).  Text on the list to follow.

Raw notes:

(Cullen presented JSEP in his role as a co-author)

Clarified how to handle various received attributes (a= line elements).
Magnus:  the rtcp attribute must be ignored?
Justin: remember that the same information is in the ICE information
Cullen:  also remember how this relates to the non-MUX
(text in draft is
o  If present, a single "a=rtcp" attribute MUST be parsed as
     specified in [RFC3605], Section 2.1, but its value is ignored.
)
Magnus:  it is not forbidden to support non-mux, but making this MUST ignore
makes Roman: it’s completely redundant Magnus: that might be true, but it is
coming across as forbidden Justin: intent is to say that you don’t need this,
because ICE handles it.  We can say “ignored” here because the information is
somewhere else.  It does not say “Must ignore”, but is ignored. Magnus:  maybe
clarifying that is enough. Cullen:  maybe what we should say is “is not used”
not MUST be ignored. Roman: splitting it into two sentence would be clearer
about where the MUST is applied. Justin:  I’ll file an issue now to make this
part clearer. Magnus: Thank you. Cullen: Any other issue for that attribute
line? Justin: I forgot to mention in the change log, is that we nailed down how
the bandwidth is handled; we’ve talked about it at multiple meetings, but the
document now has this nailed down.  Nothing should be a surprise, as it
reflects the meeting.

Revised how  attributes should be generated for bundled m= lines
(No questions)
Removed unused references
(No questions)
Remove text advocating use of unilateral PTs.
Justin:  This was a long standing issue, that came up in implementation.  It
was a pain to match up the PTs in the APT attribute.  Magnus has given some
guidance on giving RTP payload types; so we ended up removing this and being
silent on this topic

Trigger an ICE restart even if the ICE candidate policy is being
     made more strict.

(No questions)

Remove the 'public' ICE candidate policy.
Cullen: as we started the IP handling draft, this policy statement was less and
less salient, so it got removed.

Move open issues/TODOs into GitHub issues (no longer in the draft, but still
open issues that need PRs). Cullen: Text needed from the authors/contributors. 
If there is a chunk of work that needs to be done, we should have it in the
issue tracker now.

Split local/remote description accessors into current/pending.
Cullen: This doesn’t really change how things work very much, but it is a bunch
of text edits to make this clear.

Clarify a=imageattr handling.
Cullen There are some codecs that have specific of doing this, but this is a
generic method, and we clarified how it should work.

Add more detail on VoiceActivityDetection handling.

Reference draft-shieh-rtcweb-ip-handling.
Cullen: This is now a working group document, and it should be re-issued as
such.

Make it clear when an ICE restart should occur.

Resolve reference TODOs.

Remove MSID semantics.
Peter has an almost complete PR for this, incorporating the RID and Simulcast
work; this is layered on top of the transceiver work, so that PR has to land
first.  This will be done before the draft deadline for IETF 95 (PR 215 first,
then 219 next)

  o  ice-options are now at session level.
(No questions)
  o  Default RTCP mux policy is now 'require'.
Cullen: this reflects the discussion and agreements on this.

Are there new issues that were raised?
Justin: What SDP you use to re-start a DTLS session; that’s currently being
discussed in MMUSIC? Cullen:  As we shifted this, it wasn’t clear that the
language left in the full control surface in RTPSender (issue 206 is tracking
this) Justin: Issue number 230:  what’s the value for setup in a reoffer.  The
guidance we’ve given is that you should always using  ack? path or with what
you’re already using Roman:  The issue might be that with 3rd party call
control, you might be talking to a different effort Justin:  we have in that
past said that we use a subset when doing re-offer (e.g. just what was being
used before) Roman:  maybe just clarify how this works in 3rd party call
control, saying that it only needs to be used in initial offer or 3rd party
call control; in all others, re-use what was being used before. We can do that
in the DTLS draft Justin:  if you’re going to do that in the DTLS draft, we can
just point to that. Cullen: which draft Roman: mmusic-dtls-sdp

Justin: this biggest thing is when we get the transceiver landed, then the
rid/simulcast stuff.  Those will be landed early next week and we’ll emit a new
draft

Harald: and the other two PRs?  One is waiting on the transceiver stuff to
land, and 136 needs to be re-cast (adjusted title during the meeting).