address portability

["Steven M. Bellovin" <smb@research.att.com>]

At 09:12 AM 7/15/98 -0700, Patrick Greenwell wrote:
>
>Further, I would love to have a truly permanent, portable IP address(es). 
>I imagine lots of folks would. 

Right.  Lots of people would.  The problem is that at present, it's
simply not technically feasible.  Portable IP addresses -- even on a
corporate scale, where the units routed are networks, not hosts -- would
imply a very large increase in the size of the global routing table.
And that's problematic, partly because of the memory requirements -- routers
don't use the SIMMs you buy at your corner computer store -- but because
of the computational complexity of the routing table calculations.  And
that issue, in turn, is exacerbated by the desire for the Internet to
respond quickly to changes in topology, which in turn means that the
calculations need to be done rather frequently.

It may be that new algorithms and techniques will make this point moot.
For now, the best approach is to adopt technologies that make addresses
less visible and less permanent, such as IPv6 with automatic renumbering.
Or perhaps IPv6's large address space will let us experiment with other
techniques, such as geographical addressing and metropolitan area
exchanges.  For now, though, I think we have to live with topology-based
addressing -- not because anyone wants to, but because the technology
doesn't exist to do otherwise.

I should note that this issue, unlike most of the others, is almost
purely technical.  If anyone has any *detailed, technical* refutations
-- and not simply assertions that topology-based addresses are wrong,
or a bare claim that the problem is solved -- I'd be interested in
hearing it.  You're right, of course, that today it is an important
issue.  But I for one don't know how to solve it, nor to my knowledge
do the IETFers I respect.