[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Ecrit] FW: FW: I-D Action:draft-rosen-ecrit-premature-disconnect-rqmts-02.txt



At the risk of backward engineering, I don't find a requirement at the moment that would lead to this being adopted as a solution.

I believe I made comments to the effect that we need to deal with compatibility in my review.

Keith

> The current proposed mechanism is that you negotiate the 
> feature with a full
> 3 way handshake.  Then, normally, instead of termination, you 
> signal hookstate.  If you get stuck (the signaling of 
> hookstate doesn't complete, or the override is triggered), 
> you terminate.  The PSAP has a way to alert.
> All that is required to make this work in complex networks is 
> that the negotiation mechanism and the hookstate signaling 
> mechanism is transparent to networks and devices other than 
> the UAS and the UAC.  If either end really does send BYE, the 
> call terminates.   

> -----Original Message-----
> From: ecrit-bounces at ietf.org [mailto:ecrit-bounces at ietf.org] 
> On Behalf Of Brian Rosen
> Sent: Tuesday, January 13, 2009 6:30 PM
> To: 'Ted Hardie'; 'ECRIT'
> Subject: Re: [Ecrit] FW: FW: I-D 
> Action:draft-rosen-ecrit-premature-disconnect-rqmts-02.txt
> 
> I'm trying to solve a problem where the user is doing the 
> wrong thing.  By definition, his opinion is that he should 
> terminate the call.  "Negotiating"
> with the user is not possible.  We can do something and 
> provide an override, but it doesn't work to negotiate with the user.
> 
> I do understand the "intruder" problem.  I'd like to be able 
> to solve that in the best possible way.  If the user really 
> knew what was happening, they would pull the battery out of 
> the device, because any phone can be called any time.  That's 
> not a reasonable response, and I understand that.  We have a 
> balance to strike.  Of course the override could and probably 
> should be proactive (that is, if you should be able to invoke 
> the override before you terminate).  That is also not a 
> complete answer.  It seems to me that the incidence of bad 
> judgment causing serious problems is so much more common than 
> the intruder's shouldn't hear alerting that we ought to 
> compromise in a way that facilitates the former rather than 
> the latter.  Although I'm not entirely comfortable taking 
> that position, I'm a heck of a lot more comfortable with that 
> then any other proposed solution.
> 
> I'm not at all worried about the "least surprise" issue with 
> people coming in from other jurisdictions.  Surprise is the 
> most common reaction to people in the jurisdictions it's been 
> implemented in, mostly because you don't make emergency calls 
> very often.  Most Canadians don't know that the PSAP can 
> control termination on wireline calls.
> 
> The current proposed mechanism is that you negotiate the 
> feature with a full
> 3 way handshake.  Then, normally, instead of termination, you 
> signal hookstate.  If you get stuck (the signaling of 
> hookstate doesn't complete, or the override is triggered), 
> you terminate.  The PSAP has a way to alert.
> All that is required to make this work in complex networks is 
> that the negotiation mechanism and the hookstate signaling 
> mechanism is transparent to networks and devices other than 
> the UAS and the UAC.  If either end really does send BYE, the 
> call terminates.  
> 
> Brian
> 
> > -----Original Message-----
> > From: Ted Hardie [mailto:hardie at qualcomm.com]
> > Sent: Friday, January 09, 2009 5:45 PM
> > To: Brian Rosen; 'ECRIT'
> > Subject: Re: [Ecrit] FW: FW: I-D Action:draft-rosen-ecrit-premature-
> > disconnect-rqmts-02.txt
> > 
> > > > But the system we've put together here allows for someone in
> > Toronto
> > >> whose only working phone happens to tunnel back to Germany to
> > determine
> > >> the correct PSAP for their location and make an emergency call.  
> > >> How does that phone (or that network) treat the requirements for 
> > >> CPH?
> > >There is no fundamental change in the social context.
> > >You still have
> > >stressed people making poor judgments on when to end a call, and a
> > trained
> > >PSAP call taker better able to determine when to end the call.  The
> > networks
> > >are, certainly, more complex.  The requirements, and the proposed
> > mechanisms
> > >have always dealt with the circumstance you are describing 
> adequately:
> > the
> > >feature is enabled call by call at the PSAP and works 
> across complex 
> > >networks such as the ones you describe.
> > 
> > I believe that the mechanisms you've described so far would 
> only work 
> > across complex networks provided every network in the world 
> was aware 
> > of and obedient to this signalling.  The "don't tear down flows"
> > requirement
> > is pretty much not how things work in many networks now, and this 
> > would require all networks now to watch for signalling and 
> keep state 
> > in a way they do not.
> > 
> > Could we get there?  With enough will, certainly.  But 
> "works across 
> > complex networks" does not seem to describe the current 
> state of play.  
> > It is possible that I am misunderstanding the mechanisms proposed.  
> > That wouldn't be unlikely and would probably be at least partly my 
> > fault, since I've been among those asking for the 
> requirements to be 
> > worked out prior to the mechanisms being developed.  Sorry, if so.
> > 
> > > >
> > > >
> > > > Perhaps by making a requirement that there is a need 
> for agreement
> > >> here between the human user and the PSAP on who should have this 
> > >> control?  Since this isn't the same in all jurisdictions, you can
> > have
> > >> a mismatch both ways.
> > 
> > 
> > 
> > >But there isn't a negotiation between the caller and the 
> PSAP.  The 
> > >requirement we have is that the PSAP asserts control, and the user 
> > >can override.  That is described.  To implement that, I think 
> > >signaling
> > needs
> > >negotiation (both so the caller side knows what to do, and the PSAP
> > knows
> > >whether it's going to be able to assert control).
> > 
> > And here is the crux of the matter.  The PSAPs want to 
> assert control.
> > That will violate the principle of least surprise for at least some 
> > users who come from different jurisdictions or whose phones using 
> > other technologies do not behave this way, and it will not 
> be desired 
> > by others even if they aren't surprised by it .
> > > >
> > >You are stuck on the possibility the user actually does know better
> > than the
> > >PSAP what is happening.
> > 
> > I see a tension here you do not.  The user should always be 
> in control 
> > of the device, unless they surrender that control voluntarily.  The 
> > extreme example of this is the one Marc mentioned:  no one 
> else should 
> > be able to brick your phone in an emergency.  The other 
> examples given 
> > have been knocked back as unlikely (the abuser/home 
> intruder example, 
> > where the hang up *must* happen,  the anonymous tip example, etc.).
> > Maybe those are unlikely in the world these PSAPs serve. But the 
> > possibility of abuse here is also pretty scary, even if it abuse by 
> > authorities in jurisdictions different from the current 
> users of the 
> > related feature.
> > 
> > We have to get this right if we do it, and that means we 
> have to treat 
> > the desires of both parties.  From my way of thinking that 
> implies a 
> > much more negotiated approach than the "PSAP desires control this"
> > tone of the proposals to date. Absent a willingness to do 
> that, I will 
> > move on to the position "this is a bad idea, let's not do it".
> > You may well get rough consensus over my objections, but I would 
> > rather than you heard them and took them seriously.
> > 
> > If I have not said so lately, let me repeat:  thanks for 
> your work on 
> > this.  Please understand my comments are not meant to 
> frustrate you or 
> > the work; they are meant to help get it right.
> > 
> > 					Ted
> 
> _______________________________________________
> Ecrit mailing list
> Ecrit at ietf.org
> https://www.ietf.org/mailman/listinfo/ecrit
> 
_______________________________________________
Ecrit mailing list
Ecrit at ietf.org
https://www.ietf.org/mailman/listinfo/ecrit