Re: [MMUSIC] Trickle ICE comments [Re: Trickle ICE update and a new SIP usage draft.]

[Emil Ivov <[e-mail]>]

Hey all,

On 07.02.13, 11:58, Ted Hardie wrote:
> On Thu, Feb 7, 2013 at 11:53 AM, Matthew Kaufman <[e-mail]> wrote:
>> On 2/7/2013 8:18 AM, Ted Hardie wrote:
>>>
>>>
>>> It would terminate it, but not prematurely.  The client sends
>>> a=end-of-candidates whenever it's done sending candidates; it just
>>> allows you to send that indicator if you already had the full list on
>>> the first use.  So if it did the blocking version of candidate
>>> gathering, it sends that as the indicator that it is done.
>>
>>
>> Isn't part of the point of trickle-ICE that you potentially are never done
>> sending candidates? (Interface and connectivity changes and the like,
>> avoiding a complete session restart by testing and switching to new
>> candidates on the fly)
> 
> You could use it that way, but it supports modes that allow you to
> claim you are done.  The "send host candidates first, then trickle
> server reflexive and call it done" is one called out in the draft.
> Since you can claim you are done, getting the semantics of vanilla ICE
> seems easy enough.

Yup that's the idea. The whole point of having end-of-candidates
indications is to basically say: I am not done till I say I am done (and
I really am when I say I am).

Emil
> 
> Just my two cents,
> 
> Ted


-- 
https://jitsi.org