[6bone] non-global address space for IXs (was: 2001:478:: as /48)

[Stephen Stuart]

> On Mon, 8 September 2003 14:48:37 -0700, Stephen Stuart wrote:
> > If you've read the page and have questions, I'm happy to try to answer
> 
> Ok, why can ppl only do v6 transit (or rather: allowed to
> carry it, more or less, as transit is difficult to say as
> too many ppl distribute routes in v6 regardless) for the
> node in PAIX? After all we are speaking about IPv6, not much
> traffic or damage there, I think.

My position in the internal discussion has been that we don't
differentiate between IPv4 and IPv6 with respect to policy. While
today it is "just IPv6," treating them differently would leave us with
disparate policy to unify when the time comes that it's not "just
IPv6." My desire is to start clean and keep clean, with one policy and
as few reasons for exceptions as possible.

> > them. If you discussed the page in detail with Joe, though, I don't
> > know that I'd be able to supply much in the way of additional
> > information.
> 
> He just said no and mentioned some internal discussion. I'm
> merely curious as for the reasoning and implications about
> the various nodes and what is good or bad. I did not want to
> step on other ppl toes though which I seem to have done.
> CPU/ query load can't be the issue, it's IPv6 we speak of...

For normal query load, that's certainly the case. When (not if, when)
the IPv6-based attack comes, it is still the case that we want to have
the brunt of the attack borne by the PAO1/SFO2 cluster, with
sinking toward Local Nodes based on the routing policy that we
specify. Dealing with disparate routing policies in that situation
does not represent (to me, at least) the best operational practice
that we can follow.

As you say, people are much less discerning regarding IPv6 transit. To
me, that's a strong argument in favor of the no-export behavior to
increase the chances that I'll get the behavior that I want.

As Joe said, though, the internal discussion is still going on.

Stephen