Re: LLDP-MED and Phone BCP (Re: [Ecrit] New phonebcpdraft wassubmitted)

[Henning Schulzrinne <hgs at cs.columbia.edu>]

One of Brian's favorite topics and a limitation of L7 protocols:  
VPNs. In those cases, you are unlikely to get to the right L7 server  
or get it the right IP address. In those cases, lower-layer  
solutions, such as LLDP-MED and DHCP are likely to be more reliable.

Thus, I think as important as picking a list, is to also suggest the  
sequence of things to try and whether to keep trying if you got an  
answer (which may be wrong). In other words, if you got a location  
via DHCP, should you also ask via L7?


On Jul 12, 2006, at 10:25 AM, Ted Hardie wrote:

> At 4:09 PM +0200 7/12/06, Stastny Richard wrote:
> > > You need a list, where one end implements ALL and the other end  
> > > implements "at least one of" to get interoperability.  There is  
> > > no >other way.
> >
> > Fully agree. A device needs to implement ALL possible mechanisms,  
> > there is no other way, since you never know where a
> > device is attached to. Period
> >
> > Richard
>
> My agreement with this depends a lot on what "possible" means.  If  
> someone defines a pppoe mechanisms, implementing it in a phone that  
> doesn't support ethernet or pppoe is  not useful.  If you define  
> that as "not a possible mechanism for that phone", then we're  
> probably okay.  But that sounds to me closer to what Barbara is  
> saying than what Brian has said.
>
> I also believe we need to see the text Brian mentioned in the room  
> yesterday for self-configuration.  For some devices, self- 
> configuration will be more accurate than the network-provided  
> location received by many of these methods.  But putting that into  
> the list of mechanisms has some pretty interesting results.
>
> Last, the l7 question has two issues.  First, if you have IP  
> connectivity, will you be able to
> reach an l7 server?  The answer to that had better be yes, since  
> reaching one and reaching a sip service have pretty similar  
> properties.   The other question is: will the l7 server be able to  
> identify 1) you and 2) where you.  I take it that Brian's primary  
> concern is that it  will not be possible to identify a client to  
> the l7 server in ways  that will work reliably in all  
> environments.  Until we get a little deeper I can't say, but I do  
> have to question whether the effort to get to that point (so that  
> there is single fallback mechanism) is really worse than every  
> client implementing everything, where everything is a potentially  
> growing list.
>
> 				Ted
>
> _______________________________________________
> Ecrit mailing list
> Ecrit at ietf.org
> https://www1.ietf.org/mailman/listinfo/ecrit


_______________________________________________
Ecrit mailing list
Ecrit at ietf.org
https://www1.ietf.org/mailman/listinfo/ecrit