[ppml] PIv6 for legacy holders (/w RSA + efficient use)

Paul Vixie paul at vix.com
Mon Jul 30 02:05:03 EDT 2007

> You shook me from my sleep with "NAT". I live in the country where
> hack IT people feel NAT is the saving grace of the world - sadly there
> is nothing worse than having to renumber your private network because
> it conflicts with the subnet that your provider has dished out.

i refer you to ipv6 ula, ipv6 ula central, and ipv6 ula global.

> > it's a damned shame that IPv6 doesn't include a better transition
> > method.
> I can't really conceive a way in which it would include a better 
> transition method other than being able to have both at the same time.

that's what DEC did in the VAX for the first few years.  (to execute PDP11
opcodes.)  DEC later failed, but it wasn't because they got this part wrong.

the V4/V6 transition thinking as i heard and participated it in was that
every V6 node would also present V4 to its applications, either via native,
or via proxy.  the V6/V4 proxy setting would be like a default route, you'd
leard it from your DHCP server or via router solicitation/advertisement/etc.
the thing you were told by your proxy to use at your end of what amounted
to a V4-in-V6 tunnel could either be RFC 1918 or native.  so it would be
possible to run V4-only apps in a V6-only enterprise, so long as that
enterprise was connected to a dual stack core or had its own upstream proxy.

seemed like a really good balance of cost:benefit to me, since i could see
(in 1995 or so, this was) that the cost of not having seamless transition
would be huge, vs. the complexity cost of putting this logic into every node.

> The real problem with its transition method is where silly men in black
> suits think that IPv6 should be some kind of alternate profit area that is
> marketable as something other than IPv4, whereas it should be more a case of
> "This is the new standard, provision it or quit now and go sell stationary".

well, maybe so, but i think the reason V6 is mostly ignored today is that it
was sent out as "let's build a brand new internet having only a tenuous 
connection to the old one" rather than "let's add more address space to the
internet we already have."  the difference may be too subtle.  i hope not.

> My question is why it wasn't possible to learn from the NCP to TCP switch
> already performed in the internet history. I guess the issues are
> fundamentally different though.

different people, different times, and a lot more money and nodes in the mix.

More information about the ARIN-PPML mailing list