Date: Thu, 28 Oct 1999 17:54:48 -0400 From: Rob Austein Subject: IPv6, Nimrod, NATs, Evolution, Environmentalism, Politics, Vorlons, a little Zen, and the Art of Internet Maintenance A friend wrote to me I still don't like IPv6 technically.... This comment prompted a long train of thought that just may be worth writing down and possibly even publishing someplace eventually. Then again, it may be utter balderdash, but what the heck. Like my friend, I never really liked IPv6 very much. Seemed like a pretty silly outcome from all the work that went into the whole IPng effort. A bit of second system syndrome, a little cleanup of IPv4 warts, yada yada yada, but at a fundamental level IPv6 is just IPv4 all over again with larger values for some of the defining constants. At a gut level, it feels like the GOSIP situation all over again, where we were asked to believe that it was worth switching from TCP/IP to TP4/CLNP, and we concluded that it was a giant leap sideways: far too much effort for something that makes no significant forward progress. Let the short sighted fools waste their time on IPv6, I thought, we'll go do Nimrod and show 'em how wrong they were. As you have no doubt gathered, my thinking has changed somewhat. Let's skip over the lack of real progress on Nimrod, and assume that enough thrust would put Nimrod in orbit. The whole discussion about Nimrod vs IPv6 vs all the other IPng candidates misses a fundamental point, which I only gradually realized, and that only recently. The real question at this point in the Internet's evolution isn't what the best architecture for the Internet's future would be. The real question is whether there's going to be any architecture at all. If you will indulge a brief digression into philosophy, the whole model of orderly progress towards a better world is very much derived from the whole Judeo-Christian model of a plan and a purpose that starts with the Creation and ends, well, that's between you and your rabbi. Read Teilhard de Chardin for a detailed philosophy built on this model. It's an appealing vision, and arguably it's the reason why the Western mindset dominates the planet on which we live. One certainly hopes that any large engineering project has a plan and a purpose. But the Internet is no longer exactly an engineering project, and we old-timers have a tendency to forget that. At this point the Internet more closely resembles an ecosystem, with lots of independent entities doing their own thing for their own reasons, and giving about as much thought to the effects of their actions on the shared network environment as they give to the effects of their actions on the physical environment. The proliferation of NAT boxes with all their attendant problems is an example of this syndrome. The original version of HTTP was another. No malice here, just local solutions that seem to work fine at first but don't scale well. So at this point I see the question as all too simple: are we going to have an Internet architecture or not? People who aren't in positions of responsibility can punt this question. Individual members of the IAB can punt this question by leaving the IAB. The IAB as an institution can not escape this question; failure to answer is itself an answer. I still don't like IPv6 very much, but it's the candidate from the "Yes, we do want an architecture" party. Its main strength is its similarity to IPv4, which should, in theory, make it relatively easy to deploy, since it doesn't really involve anything radically new, just a lot of software upgrades. Even so, it's behind schedule and the outcome is in doubt. I use the word "candidate" deliberately: if you ignore the technical issues (an area where reasonable people can and do disagree with each other's conclusions), what you're left with looks an awful lot like politics. The IPng effort was the primary, and it picked a candidate. The primary was ugly, and the candidate that made it to the nomination wasn't the one I would have chosen, but that's normal. The primary is over now. If you believe in the party, it's time to put the primary behind us and work for the nominee, because that's our best hope of beating the other party. To complete the picture, consider the really big companies who find the whole open standards process to be an annoyance, and who wouldn't really mind seeing the open standards process totally fail to cope with the continued growth of the Internet. Their winning scenario is really just the ecosystem model again, writ large: basicly what you get if they win is a monoculture, where a small number of players control the key technology and everybody else has to play it their way. Pick your favorite potential villain (it's a target-rich environment), but also remember that in almost every case there are good people working at each of those companies today, some of whom fight essentially this battle daily within their own company. Some win, some lose. So what we have here is your basic epic battle between the forces of Order and the forces of Chaos. While my knee-jerk sympathies are usually with the forces of Chaos, I kind of like the Internet that the forces of Order built, I think it's in danger, and I'd miss it if we lost it. Anyway, the forces of Order are vastly outnumbered, and I never could resist rooting for the underdog ("Go, Cubs, Go!"). Their own emotional immaturity in extending the concept to the other team notwithstanding, the Vorlons were basicly right about the need to play nicely with the other kids if you're going to accomplish anything constructive. Oh yeah, I promised that there'd be some Zen, here, didn't I? It's a bit long to type in, and it's not exactly Zen, but see Sam's sermon to the monks of the Buddha after the death of Lord Mara at the beginning of Roger Zelazny's "Lord of Light." Probably the most inspired piece of metaphysical bullshit in a good cause that I've ever read.