Chris Wedgwood
Its gaining momentum every day. I first connected to the 6bone in '96, there were about 30 or 40 sites then, now there are hundreds with probably thousands of hosts down-stream.
Questions: Where do I buy an IPv6 connection? Where do I get IPv6 address space from? Oh, and I don't want to talk IPv4 to anyone, just IPv6, but I'd better be able to talk to the whole Internet. Frankly, the IPv6 situation fills me with a sense of deja vu. It feels a lot like the deployment of OSI in the early 90s, with lots of things on the critical path being re-invented unnecessarily. (I don't think OSI ever managed to get its bloated Virtual Terminal spec out the door before the momentum for OSI/GOSIP fell away, back when the Internet community had just stopped worrying about it and and made terminal emulation Somebody Else's Problem in Telnet.) To me, IPv6 has failed to be widely deployed for two reasons. One of them is sheer bloody-minded arrogance on the part of the IETF protocol gods -- a quite sensible proposal, TUBA, was put forward as one of the four IPng proposals, which involved basically leaving TCP and UDP alone, and running them over OSI CLNP packets. If pursued, we wouldn't be having this discussion -- we'd be running all our current services over TUBA packets instead of IP. The real beauty of TUBA was that at the time there had been enough push for OSI from the government community that most infrastructure actually supported it (or at least could turn it on). If they'd released TUBA as the replacement for IP back in 1995-96, we'd have just turned it on, dealt with the issues and (mostly) turned IPv4 off about 1998. The scale of the migration problem in that timeframe was a very tiny fraction of what it is now. But TUBA was based on OSI, and OSI was The Enemy. I watched people like Tony Li (cisco) banging their heads on people who'd never had anything to do with packet switching who would say, "but OSI is slow", and Tony would say "we can switch it as fast as IP", "but only under special circumstances", "all circumstances", "can't possibly", "we can switch it *as* *fast*", "can not" ad nauseum. Some of these people had no idea about protocol switching hardware and software; as far as I could make out, many who voted for what became IPv6 had simply assumed that OSI's variable length addresses were slower than fixed length addresses, without even bothering to ask the people who made routers for a living if this was genuinely a problem. (Techy aside: variable lengths aren't a problem. OSI addresses have address families, so you can assume particular rules for particular address families -- the addresses are no more "variable length" than CIDR prefixes, or for that matter, classful IP addressing.) The other reason for IPv6's failure was that the IPv6 folks insisted that authentication, encryption, address assignment and other stuff had to be done in the base protocol -- and deployment couldn't start until these had been done. In IPv4 land, we do all of these things now in higher level protocols (although some of them aren't done well, mainly due to IETF myopia -- as far as I can make out, there are a bunch of IETFers who don't believe that NAT could ever possibly work and therefore doesn't need to be considered <sigh>, IPSEC being a case in point), such as IPSEC, SSL, DHCP and friends. So I'm not optimistic about IPv6. I think an opportunity to make an incremental but rapid change to solve the immediate problems was squandered by bad cases of Not Invented Here syndrome and second-system effect. I mean, I can run DECnet by tunnelling it through IP, but I don't think I want to. (And once apon a time, some of the world's biggest data networks ran DECnet, although 16 bits of address space divided into 1024 nodes in 63 areas was a tad limiting -- DEC's own Easynet had "hidden areas" that looked suspiciously like the use of private IP address space, mumbleteen years ago.) And I don't think I want to run IPv6 right now either. -- don --------- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog