Re: [RRG] Consequences of no renumbering...

[Scott Brim <swb@employees.org>]

Excerpts from Tony Li on Wed, Sep 10, 2008 11:00:15PM -0400:
> 
> Hi all,
> 
> So in thinking more about our recent consensus on renumbering, it seems to
> me that this helps us prune the solution tree a bit.  In particular:
> 
> - For the entire class of 'transport' solutions that we've discussed, it
> seems like renumbering would always be required for these solutions.  All of
> the obvious transport protocol changes would utilize multiple PA addresses,
> and since changing PA addresses would result in a renumbering event, it
> seems like these solutions should be avoided.
> 
> - For the map-and-encap solutions, there seems like a similar issue.  If we
> look at the current LISP transition plan, there is currently a requirement
> for sites to renumber once to get into an aggregateable EID space.  
> 
> If renumbering is not required, then EID space doesn't aggregate.  If
> transition boxes (PTRs) advertised EID space into legacy routing, then it
> would imply that there wouldn't be any reduction in prefix count until
> transition was wholly complete.  That doesn't seem very practical.
> 
> I'm not ready to say that there aren't transition schemes that could get
> around this, but these are the issues that I'm seeing.
> 
> Comments?

Sorry for not having replied on your previous proposed summary of the
consensus before this.  As I believe Robin pointed out several days
ago, I don't think we agreed on simply "no renumbering".  That's too
black-and-white.  The consensus I saw was that we should not adopt an
architecture that assumes site renumbering is easy.  Based on what
network operators have said, once seems okay, and even "rarely" might
be acceptable if other benefits outweigh the cost.

As to whether sites must renumber (at least) once for map-and-encap
systems (actually just the map part), we are still uncovering new
ideas on how to approach all that so I'm not resigned yet.
Renumbering is not required for a mapping system to work, since local
numbers are decoupled from overall topological location.  It might be
required for interworking in the initial phases of deployment.  

But the main point is that afaict the consensus was not just "no
renumbering".  Rather, I think it was that we should not choose an
architecture that assumes it is easy.

Scott

--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg