Re: [Ecrit] IETF ECRIT Design Team on Premature Call Termination

[Henning Schulzrinne <[e-mail]>]

I think there's a subtle difference between "accidental" and  
"premature" call termination. The former is easy to define: the caller  
accidentally disconnected, either because of hitting the wrong button  
or some technical problem, and would have preferred to stay on line.  
The second is harder to define, since both sides may have a different  
opinion as to whether the disconnect is "premature". Accidental  
disconnects are also premature, but not the other way around. The  
difference is the set where the PSAP wants to continue to talk, but  
the caller doesn't.

We can prevent accidental disconnects largely with UI features, such  
as requiring user confirmation for hanging up.

I think we can also prevent most legitimate premature terminations of  
the "caller doesn't realize the call taker wants to continue to talk,  
but has no objection" with appropriate UI design. A simple modal  
dialogue with

"Please wait until the emergency call taker disconnects. Do you really  
want to terminate the call now?"

before sending the BYE will presumably deal with most cases.

The correct behavior depends a bit on the device. For example, an  
iPhone has different possibilities than a traditional place-receiver- 
on-hook phone (although that type of phone could conceivably switch to  
speakerphone mode until the user confirms the disconnect, with some  
kind of alerting to prevent eavesdropping).

None of this requires more than recognizing that this is an emergency  
call. This is tricky for two cases: proxy-recognized calls and return  
calls.

Henning

On Nov 6, 2008, at 2:51 PM, Randall Gellens wrote:

> At 1:57 PM -0400 10/8/08, Brian Rosen wrote:
>
>> Sure, UI is important here.  Some text on multiline might be  
>> appropriate
>> that is all about UI and But the IETF problem is the signaling: no  
>> one is
>> arguing that it is sometimes the right thing to let the user  
>> disconnect.  So
>> it's not an absolute, "don't send BYE" or disable the button on the  
>> UI.
>> There is negotiation of the feature.
>
> I think the issue of PSAP control over termination is both  
> signalling and UI.  It can't be UI-only, because it needs to be  
> optionally invoked by the PSAP.  Presumably, PSAPs only want to  
> invoke it when the call is being handled by a call taker, not when  
> it's in a call queue or IVR, so there definitely needs to be  
> signalling.  Also, the PSAP probably needs to be able to initiate  
> the signalling regardless of whether it is the called or calling  
> party.  Consider situations where someone places a call, and the  
> PSAP needs to call back for whatever reason (and there are multiple  
> possible reasons why the PSAP might need to call back, even in the  
> presence of this feature).
>
> Given this, it probably makes sense that control over call  
> termination be negotiated by SIP signalling that can be sent during  
> the call by either end.  Presumably, the user's device would have  
> some criteria for accepting such a request to cede termination  
> control (since you'd want this to be accepted when made by a PSAP  
> but not when made by a telemarketer).  Once the user's device  
> accepted such a request, presumably it would not send a BYE, and the  
> UI would indicate the situation to the user, hopefully taking full  
> advantage of the UI richness afforded by modern devices.   
> Presumably, a PSAP wanting to invoke the feature would make the  
> request when the call taker came on the line, thus avoiding problems  
> with IVRs and call queues.
>
> -- 
> Randall Gellens
> Opinions are personal;    facts are suspect;    I speak for myself  
> only
> -------------- Randomly-selected tag: ---------------
> Procedures!  Of course, that's what you'd use with a *sophisticated*
> machine like this.  Procedures.    --'Ceaser Smith' in _Hot_Millions_
>

_______________________________________________
Ecrit mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/ecrit