Re: [MMUSIC] BUNDLE TEXT: ICE Procedures (18.6)

[Harald Alvestrand <[e-mail]>]

On 06/18/2013 12:17 PM, Christer Holmberg wrote:
> Hi Harald,
>
> 8.3.  ICE Connectivity Checks
>   
>     Once a BUNDLE address has been selected for the Offerer and the
>     Answerer, it is RECOMMENDED to perform ICE connectivity checks and
>     keep-alives [RFC5245] for the whole BUNDLE group, rather than for
>     each individual "m=" line in the BUNDLE group.
>
>> I have a bit of trouble parsing this last paragraph.
>>
>> There are two possibilities:
>>
>> - The offer/answer is incomplete, so we don't know whether BUNDLE will be turned on or not. (This can happen only in updates - you can't
>> do ICE checks before choosing the BUNDLE address in the initial exchange, since you don't know the remote candidates)
>>
>> - The offer/answer is complete, and both parties know the BUNDLE address
>>
>> In the last state, it's impossible to do anything BUT check connectivity for the whole BUNDLE group; there is nothing else to check for. So RECOMMENDED doesn't make sense; it's not even a MUST, it's just the way things work.
> I have no problem with adding MUST.

I'd prefer that. Or even change "it is RECOMMENDED to perform ICE 
connectivity checks" to "ICE connectivity checks are performed". If 
there's only one possible way to do it, it seems superfluous to use 
words that look like there's an alternative.

>
>> In the first state, I think it should be RECOMMENDED to delay checking until you have completed the exchange, so that you know what the BUNDLE address is. But that's not what the words seem to say.
>>
>> Could you clarify?
> Isn't that covered with the "Once a BUNDLE address has been selected"? Or, should I say "negotiated" instead?

Yes, on second reading, that's implied - you talk about my second state, 
not the first one.

>
> Or, did I misunderstand your comment?
>
> Regards,
>
> Christer
>
>