Re: [rtcweb] Open data channel issues

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

> From: rtcweb [mailto:[e-mail]] On Behalf Of Randell Jesup
> Sent: Tuesday, March 04, 2014 3:01 AM
> On 3/4/2014 2:55 AM, Michael Tuexen wrote:
> > On 04 Mar 2014, at 07:30, Randell Jesup <[e-mail]> wrote:
> >
> >> On 3/4/2014 2:00 AM, Martin Thomson wrote:
> >>> On 4 March 2014 03:38, Randell Jesup <[e-mail]> wrote:
> >>>> To take your suggestion:
> >>>> a) applications assume 16KB if ndata wasn't negotiated (per the current
> >>>> suggestion)
> >>>> b) implement ndata and EOR support in the browsers, and drop the 16KB
> limit
> >>>> when ndata is negotiated.
> >>>> All this needs is "is ndata supported by this association" exposed to
> the
> >>>> app; no sdp changes needed.  (One less dependency!)
> >>> Or, to avoid vestigal crap in the API in the end state, when you
> >>> install the 16K limitation, add an indicator that applications can
> >>> check.  Once you remove the cap, remove the indicator.
> >>>
> >>> if (dataChannel.isNotVeryGood) {
> >>>    chunkMessage(message);
> >>> } else {
> >>>    dataChannel.send(message);
> >>> }
> 
> 
> Wait, the reason why we had it be 16K as a default was that current
> impls don't have the property.  So datachannel.isntFixed (ok, better
> names can be found - but that's W3 anyways.

[Raju] 
NDATA requires support from both the ends. Is browser going to update datachannel.isntFixed/ndata after SCTP is up? Per my reading of PeerConnection onopen() won't be called before SCTP is up (however may be called before DCEP, if used, is done). So, app knows NDATA support before sending first message.

Thanks
Raju