Tunnels...

[Joe Abley]

On Mon, Feb 14, 2000 at 03:57:13PM -0500, Greg Maxwell wrote:
> On Mon, 14 Feb 2000, Bob Fink wrote:
> > At 10:39 PM 2/13/2000 -0500, Jason wrote:
> > >Is it acceptable for me to dish out tunnels as an end site, and if I do
> > >this, should I document them with tunnel attributes.
> > 
> > If you are a pNLA end-site (versue a transit) with a /48 then the only 
> > tunnels would be for parts of your own network, i.e., some other SLA ID.
> 
> Since only my home is on the 6bone right now (as opposed to my work), I'm
> not reall one to comment here but...
> 
> I don't agree, what about peering? Even end-sites can peer with other end
> sites to exchange traffic among themselves. This is useful in the face of
> v6-in-v4 tunnels causing very unoptimim paths between end-sites. This is
> also useful when they are IPv4 peers in 'real life'/ 

This discussion sounds like it might be veering off towards questions
of multi-homing, in which case a recurring argument might well ensue
(speaking from experience :)

In case there's anybody new here who has niggling doubts about how
pure aggregation is going to work with widespread multi-homing of
"end site" networks, there have been many discussions on the topic
already, which you will find in the ipng and ngtrans IETF working
group archives. I'm not saying there are answers there, just that
there are arguments there which are worth not re-hashing.

[Note that I know the topic in question was peering, and peering
between end sites does not have the same implications. Just that
when you start peering, its only a matter of time before you someone
starts to talk about backup transit arrangements.]


Joe