Re: [rtcweb] *MUST* DCEP?

["Makaraju, Maridi Raju (Raju)" <[e-mail]>]

> >>> Are you looking at more than what is documented in
> >>> [I-D.ejzak-mmusic-data-channel-sdpneg]?
> >>>
> >>> Or are you looking for an indication of this (for all streams) directly
> >>> in a=sctpmap?
> >>
> >> I'm ok as things currently stand, though the documents are confusing.
> >> The 'a=sctpmap:nnn websocket-datachannel' means that DCEP is supported.
> >>
> >> My concern was a proposal I think I heard that would make support for
> >> DCEP to be optional. In that case, IMO it is important that the O/A
> >> exchange establish whether DCEP is supported or not.
> >
> > [Raju]
> > That's the reason why I proposed to add 'protocol' parameter to a=dcmap.
> > Having this parameter at a=sctmap level may not make that much sense due
> to:
> > sctpmap is for entire association and protocol defaults to DCEP. To make
> use of external negotiation, which is per data channel stream, one would
> need additional a= line(s) (e.g. ejzak's draft) to convey subprotocol info.
> IMO adding data channel protocol option to a=dcamp, which is per
> subprotocol, makes sense.
> > [/Raju]
> 
> I have no problem with a=dcmap.
> 
> It is the 'app' parameter of the sctpmap attribute that I am talking about.
> 
> If I want to use DCEP, the dcmap attribute is irrelevant to me. What is
> it that tells me that DCEP is supported. The only datum available that
> could imply that is the 'app' parameter of sctpmap.
> 
> Right now, if app is webrtc-datachannel, that implies DCEP is MTI. That
> serves the purpose.
> 
> But as of today there was a suggestion that you could have cases where
> there is an SCTP association that supports Data Channels, but not DCEP.
> I don't know how the two would be distinguishable. Possibly by a
> different 'app' value.

[Raju] 
Yes, one option is to define a new 'app'. 
Another option is to use 'webrtc-datachannel', but then convey your own
external data channel protocol via SDP and use it for all the streams.
When you say "not DCEP", you mean it is "external" or "a different flavor
of internal DCEP"? 

[/Raju]

> 
> A related question is what is the applicability of a=dcmap. Surely it
> doesn't apply for all uses of SCTP - those that don't use data channels
> at all. Do you propose to couple this to particular 'app' values? Or is
> support for this negotiated independently of a=sctpmap?
[Raju] 
Correct. dcmap applies to SCTP data channels only. dcmap's 'protocol' parameter
allows you to override webrtc default DCEP per data channel stream.
When you are using dcmap it allows per data channel "interal(DCEP)" or 
"external(SDP based)" determination using 'protocol' parameter.
If there is a need we can allow the 'protocol' value to be extended (per IANA, may be?!
):
"internal-DCEP"
"external-SDP-based" (ejzak's draft)
"new-enhanced-internal-DCEP"
"pre-defined-stream-id-use-arrangement"

[/Raju]

Best Regards
Raju