ietf
[Top] [All Lists]

Re: Solving the right problems ...

2003-09-15 07:07:37
vinton g. cerf wrote:

I am a strong proponent of trying to find a way to create a new set of end 
identifiers that would be insensitive to the changing of IP level addresses. It 
seems to me that we would find ourselves working pretty hard to tease apart the 
current strong binding of IP and TCP (pseudoheaders etc) but it may be well 
worth the effort.


We did this in SCTP.. we did NOT place a pseudoheader in specifically so that this would not be
a problem.. especially since SCTP had to deal with multiple addresses..

For one thing, it might lead to the ability to carry TCP segments over multiple 
Source/Destination pairs between the same hosts (labeled by a single end point 
identifier each) in addition to allowing for rebinding of endpoint identifier 
and IP address.

Sounds just like what we have in SCTP as well .. at least with the SCTP ADD-IP extension :>

The rendezvous and signalling problem of concurrent motion is not unlike the challenge of TCP's simultaneous-INIT - you have to get the fixed point right to make it all work.
We actually made the simultaneious-INIT in SCTP without a fixed point.. it was a
bit tricky but it does work :> However if you want to do dual ended IP
address changing you DO need some fixed point somewhere. Someone once
made a proposal to me on how it could be done with just a temporary
state repository.. at least for the both moving and already established case.

For the moving and NOT yet established this gets harder.. the fixed point
needs to be something like a mobile IP home agent.. with a lot
less functionality.. i.e. tell me where X is... and you get back a
current location... more of a DB dip then anything...

We have other fixed points in the Internet, notably the root hint file, so 
perhaps it is not unreasonable to consider another fixed point concept to 
facilitate simultaneous rebinding of IP and endpoint identifiers. I suspect 
this gets pre
tty messy when you start to think about multicast but that's territory that also needs exploring.

Yuck.. I that one makes my head hurt a bit.. I am sure it can be done but
we would need to think about it very hard :-0

We would also want to look very carefully at the potential spoofing opportunity that rebinding would likely introduce.

This is one of the reasons the authors of ADD-IP have NOT pushed to get it done.. some more
work needs to be done on this area...

R


Vint

At 05:44 PM 9/12/2003 -0400, Keith Moore wrote:
and of course neither SCTP or TCP would be sufficient by itself.  we still
need a suitable identifier,  a way to map those into locators, and a way to
maintain those mappings.

I'm still undecided about whether it is better to modify existing transports
or to do a mobile-IP like approach.  The latter has tunneling overhead but
works for all transports and in some sense the changes are simpler.

Vint Cerf
SVP Technology Strategy
MCI
22001 Loudoun County Parkway, F2-4115
Ashburn, VA 20147
703 886 1690 (v806 1690)
703 886 0047 fax
vinton(_dot_)g(_dot_)cerf(_at_)mci(_dot_)com
www.mci.com/cerfsup




--
Randall R. Stewart
randall(_at_)stewart(_dot_)chicago(_dot_)il(_dot_)us 815-342-5222 (cell phone)