On Fri, 2002-11-22 at 09:34, Markku Savela wrote:
> 
> > Fine by me.  It's just that dealing with scopes seems to be the problem that
> > most people are complaining about rather than the existence of the addresses
> > themselves.
> 
> I cannot understand those people complaining about scopes. We will
> have
> 
>   (a) addresses that are global
>   (b) addresses that are local, with limited reach (firewalled or
>       whatever)

I'd take a step back (up?) to a higher level and say there are only
really three fundamental scopes :

1) External
2) Internal
3) Link local

Reliable External communications requires globally unique addressing.

Reliable Internal communications does not require globally unique
addressing, but does require internally unique addressing.

The current site-local model seems to me to be a too restrictive
solution to the Internal communications requirement, as it creates
artificial boundaries on the definition of Internal communications.

Further, with the current site-local model, you have to assume that
there will be addressing collisions when merging even the smallest of
two "sites" - there is a fairly good chance that most people will use
fec0:0:0:1::/64 for their first subnet.

This duplicated address space, and almost guaranteed chance of address
space collision seems to me to be the cause of a lot of the recent
arguments about site-local addressing, including problems cross site
address referrals etc, and solutions such as multi-site routers etc that
just feel to be way to complex to me. 

Pekka's solution reverses that assumption. In the majority of cases, you
won't have a duplicated (on a global scale) Internal address space,
making merging "sites" a simple as bringing up the link between the
sites, and pushing a few routes between the networks.

On the rare occasions that there is a Internal address space collision
under Pekka's model, it will (or at least should) be discovered during
the preparation for the connection of the two "sites" together. In the
case of a collision, having to renumber one of the Internal address
spaces to make it unique again is unfortunate, but it is the price of
requiring an Internal address space that is independent of your provider
supplied global address space.


Regards,
Mark.



--------------------------------------------------------------------
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]
--------------------------------------------------------------------

Reply via email to