Test for IPV6

[Robert E. Seastrom]

"Mike O'Dell" <mo at ccr.org> writes:

> I don't understand this "test" - back when NCP gave way to TCP,
> almost 30 years ago, there was such a test, but that was because
> there was only one network run by The Great and All-Powerful Oz.
> He hasn't been around for, oh, 30 years!

The idea behind this is not to scare people who have IPv4 only (as any
credible web site/email/etc will be dual stacked for the foreseeable
future) but rather to smoke out folks who appear to have working v6
but don't (for instance 6to4 that happens to anycast to a relay on
another continent).  Teredo, 6to4, and other auto-configured
transitional protocols all suffer from this weakness.  6rd was
introduced in an effort to address this problem, but has the
unfortunate side effect of being a profligate waster of space.

Geoff Huston (with whom I disagree on many things) has a nice write-up here:
http://www.potaroo.net/ispcol/2010-12/6to4fail.html

The idea is to fire a round across the help-desk bows of commercial
and residential ISPs by turning on AAAA records simultaneously for a
lot of popular stuff.  Given the timetable for IANA exhaustion (after
which the RIRs will fall like a domino train) I think this is a
splendid plan.

> You should be able to tunnel v6 though v4 over FIOS just fine.

Indeed, my v6 in my place is tunneled in from my colo at Equinix.
Works fine.

> If you mean v6-native, I have *no* intention of helping
> debug that deployment horror-show. I did my time back then when
> my IMP at LBL, IMP-34, was the first to move to MILNET which
> was TCP-only, and LBL-CSAM (VAX-11/780) was running 
> 4.1a,b,&c BSD in quick succession.

Actually, I anticipate that native v6 will be a lot less of a cluster
than the tunneling and 6rd that we are looking at today.  Only
downside is this small matter of needing forklift upgrades for the
head end equipment.  Nobody ever said this was gonna be cheap.  Sure
beats the crap out of modem pools and drying your hair in the output
air stream of a bunch of MAX TNTs...

> In fact, because LBL had one of the first class-B addresses ever
> assigned, I personally hacked address classes into the 4.1a kernel.
>
> If we'd just been smart enough to expand the address to 64 bits back
> then before "mass deployment", we'd be in a very different place now.
> But nope - they were clever enough to invent address classes, 
> but not smart enough to take the hint and NOT do that.

Yeah, there is that...

-r