Re: [rtcweb] RTCWEB Data Channel: message interleaving

[Paul Kyzivat <[e-mail]>]

On 2/6/14 3:52 AM, Michael Tuexen wrote:
> On Feb 6, 2014, at 9:38 AM, Christer Holmberg <[e-mail]> wrote:
>
>> Hi,
>>
>>> If NDATA is not supported, the message size is limited to avoid the monopolization. This limit will be signalled through SDP.
>>
>> What do you mean by "signaled through SDP"?
>>
>> Will the browser inform the JS app, using SDP, if the local and/or remote peer doesn't support NDATA?
> What will be signalled in SDP is the maximum message size the end-point can handle. This is needed
> anyway... If you don't support NDATA, you can signal a smaller value of, let's say 16KB or so.
>
> NDATA is negotiated via the SCTP setup and there is no need for another signalling channel.
>
> After the association is up, both sides know if NDATA is supported and, if I remember correctly,
> there is a JS way of making message size limits available to the application. So you can
> limit the message size if NDATA is not supported by both peers. So that is the plan:
> * If NDATA is supported by both peers, message size are only limited by reassembly
>    memory and any limit is signalled via SDP. The corresponding support will be in the
>    next revision of the SDP ID.
> * If NDATA is not supported by both peers, a message size limit applies to avoid monopolisation.

I assume that in the RFC this will just be a MUST NOT. Will that apply 
to the SCTP implementation? Or to the SCTP user?

(In a practical sense, will this be implemented in the SCTP stack, in 
the DataChannel "stack", in the browser, or in javascript?)

	Thanks,
	Paul