IPv6 multihoming

[Brandon Butterworth]

> Whether or not operators want to permit /33.../<whatever> from /32s
> is still under discussion

We probably left it a bit late already

> I've tried to push this to the RIPE routing
> working group, as this is really something "operators should agree upon"

I think it's essential to avoid repeating v4

> but so far (unless I've missed something), the routing WG hasn't been able 
> to come to consensus what they want to see as "official recommendation".

I'd say don't allow it. It has its uses but the cost is too high

Now don't may be an agreed number of bits or none at all but it needs
to be a hard, universal, limit or people will erode it, there's always
someone else willing to break the net to take your customer

The RIRs have already made it expected, we have a bunch of different
assignment sizes enough that I doubt many will filter.

Though people don't like to think of going back to classful it would
have been convenient to have a bunch of fixed size allocations we
knew everyoen filtered on. It failed in v4 as there wasn't enough, not
a problem for v6

> (Now, I'm a strong believer in aggregation, but I also recognize that
> there are cases where ISP networks are run by the same entity but 
> for some or the other reason are not connected - so they have a single
> /32 for both [or even multiple] parts, and *need* to deaggregate...)

No they don't, give them extra separate space.

For that little convenience for a few why should the entire net be at
the mercy of rampant deaggregation either eating your routers or letting
people do more specifics of your routes, some will deagg down to whatever
is the limit (/48?) to protect themselves from that.

If they're going to take a routing slot regardless you may as well give them
one with less risk

brandon