Alain, Bob and Robert,
You bring up some very good points in that it would be very good to detail the case for long-lived addresses where the proposed end to end route optimization provides the intended value.
I see only three scenarios that make sense and I have reservations about the proposed end to end solution with each:
1> a person does not wish to disclose their true location to any level of granularity both to snooping parties and to communicating parties.
reservation: It is very simple to dump the packets received at a communicating party and use a traceroute on the mobile's COA. The proposed solution gives up route optimization for this location privacy so the excuse for the feature is mute. If a person does not require location privacy then they can dynamically allocate an address from their initial point of attachment and anchored mobility can be used if they go off-domain. In this case the optimization is not that big of an issue. It does save bandwidth over skinny links back to the anchor, though.
2> a mobile router
reservation: This just seems very odd to me to have communicating hosts be told the location of every single node attached to such a device especially since the police, defense and clandestine industry is a major customer of such devices.
3> IP VPN
reservation: I'm not sure what admins are going to think about putting the security of their networks in the hands of the employee's hosts. There is currently little verbage in the draft about these scenarios. Also the case where there might be a skinny link is this VPN case. So the bandwidth saving argument for scenario 1 falls apart.
4> Every Single Node ever connected to IPv6 has a long lived address or at least a home prefix.
reservation: This is a neat idea but due to Moore's law, devices get old and chunked every two years if not sooner. This doesn't seem like a good long term solution even with the expanded address space.
Does anyone know of any other cases that make sense short of the PGP security model?
I do think route optimization of some sort of standardized and ubiquitous would add significant value and not only in terms of mobility. I'm just not sure if a completely end to end is the answer. I certainly woudn't get bent out of shape if the current proposal was mandated though as I think the ubiquity of the function could be used for other useful purposes than just mobility.
Glenn
> -----Original Message-----
> From: Alain Durand [mailto:[EMAIL PROTECTED]]
> Sent: Friday, May 31, 2002 3:31 PM
> To: Robert Elz
> Cc: Bob Hinden; IPng List
> Subject: Re: Mandating Route Optimization
>
>
>
> On Friday, May 31, 2002, at 03:31 AM, Robert Elz wrote:
>
> > Date: Thu, 30 May 2002 16:43:50 -0700
> > From: Bob Hinden <[EMAIL PROTECTED]>
> > Message-ID:
> > <[EMAIL PROTECTED]>
> >
> > | It doesn't seem right to make them non-compliant (i.e.,
> make RO a
> > MUST).
> >
> > Bob, that's completely bogus as an argument. I don't know
> enough about
> > the issues to comment on the substance, but if RO is
> something that the
> > WG feels is important for all nodes to implement, then of
> course MUST is
> > the right thing.
>
> When a new technology solves a critical problem for the operation
> of the Internet, it make sense to make it a MUST, regardless of
> the installed base.
>
> The issues here are to know if:
>
> a) RO solves a critical problem to the operation of the Internet
> b) RO is the right approach to the problem.
>
> If I can be convinced of a), b) is another story.
> Years ago, the Home Address option was defined and
> presented as _the_ thing that will solve _the_ problem.
> Then we discovered that there were security concerns,
> and a whole new approach, much more complex is now
> presented.
> How confident are we that this time, this is the right thing?
> Has it been tried in large scale environment? Are we
> sure it solves enough security concerns that another
> version of RO will not come next year to fix it?
>
> As some said early, there are a number of implementations
> out there today. It is not that it is not possible to change anything
> at this point, it is that there is a need for a certain level of
> confidence
> that RO, as defined today, is the right thing and will not change.
>
> - Alain.
>
> --------------------------------------------------------------------
> IETF IPng Working Group Mailing List
> IPng Home Page: http://playground.sun.com/ipng
> FTP archive: ftp://playground.sun.com/pub/ipng
> Direct all administrative requests to [EMAIL PROTECTED]
> --------------------------------------------------------------------
>
