Re: [splices] SPLICES IETF-80 meeting minutes and start of ML discussion

[Paul Kyzivat <[e-mail]>]

inline

On 4/27/2011 2:45 PM, Salvatore Loreto wrote:
> On 4/16/11 2:23 AM, Paul Kyzivat wrote:
>> Sal,
>>
>> On 4/15/2011 2:02 PM, Salvatore Loreto wrote:
>> ...
>>> I am interested in scenarios where both ends may be impacted by the
>>> implementation
>>> of the new mechanism to be developed;
>>> of course in the case an impact is required, I do think the design
>>> should consider a fall back for the case the other
>>> peer does not implement it (e.g. a legacy UA)
>> ...
>>> work of a loosely-coupled coordination scenario,
>>> where no single device needs to remain in the session for its entire
>>> duration
>>> and no single device needs to act as master...
>>> is still something that I personally find interesting to be investigated
>> I still don't *get* it.
>>
>> Can you lay out a scenario/use case/example of what you have in mind?
>> - describe to moving parts
>> - what interactions they have to make the case work
>> - what capabilities each has/expects of others
>> - if parts of the coordination depend on special behavior of
>> *people* rather than automata, explain that too.
>>
>> Nothing complicated - just very high level.
>
> what I have in mind is a scenario where
> a user can make or answer a Call Using Two or more separate devices;
> the devices can happen to be in the same network or in different area
> networks,
> however I don't want any of the devices to be a single point of failure
> for the call:
> that means that any device, involved in the side of the call
> (the one is using different devices to handle the call), can go out of
> coverage, can restart etc.etc.
> without having any impact on the call as such.

Well, maybe a little more detail that that :-)

The above seems to be of the "and then magic happens" school.

I presume you have in mind that the *other* party manages two separate 
dialogs and coordinates them.

But who has the intelligence to establish the multi-device call?
- know that more than two devices should be used
- know how to reach the additional device
- know enough about the existing call to splice it in

	Thanks,
	Paul

P.S. I think a variant of the 3pcc approach we are discussing in another 
thread could potentially be used in your scenario too, if the main point 
is to manage where the failure point is.

For instance, if Alice and Bob are talking, and Alice wants to add in a 
local display device, it could send a REFER to the display device, 
asking it to send and INVITE/Join to Bob. This is not so straightforward 
- there are many details to work out. But I find it no less plausible 
than any other yet-to-be-devised scheme.