Photo by dhepnar
I’ve been following the story of IPv6 for a while now, and I’ve been following it like I believe most system administrators have been. Which is the standard waiting-for-when-I-have-to-drop-everything-and-get-it-implemented way. I’m sure that this attitude makes some people unhappy, after-all we’re running out of IPv4 addresses and I’m keeping us from solving the problem. If we all took the time now to get things switched over then we could avoid IP Addressageddon.
Well, it’s not as bad as all that. In my, admittedly limited, experience I have learned a few things which makes me a little less worried. Less worried, but still a bit worried.
1. The death of IPv4 has been greatly exaggerated.
Available IPv4 addresses were supposed to run out this year, but that date has been pushed back to 2011 despite a (somewhat) mad rush on addresses. The date will probably be pushed back again, but each time it does, the cost of not transitioning will increase. No one wants to be left without a chair when the music stops, but on the flip side, no one wants to spend time or money now when there still seem to be quite a few chairs left.
Personally, I think that IPv4 will be with us long until after the IPv6 transition is essentially complete and that some form of technology will keep it alive in the same way that there are still people on dial-up. In the same vein, though, there are good reasons to get off of dial-up.
2. Vendors are already there, so there’s no excuse!
One thing that I hear from IPv6 advocates (zealots?) is that vendors are supporting IPv6 in droves and so it should be simple to move our networks over. IPv6 has been in Windows, Linux, various UNIX, and OS X for years (with hardware for about as long) what’s the hold-up?
Well, implementing IPv6 at the vendor level is (not to put too fine a point on it) a helluva lot easier than at the system administrator level. It’s nice having all of the pieces of the puzzle, but it’s the job of putting them together that is the hard work.
3. IPv4 mapping could have been better.
I think that the IPv6 standards committee dropped the ball here. I get the feeling (and this is based solely on my impression, I haven’t done enough research to verify) that IPv6 was designed to be as minimally compatible with IPv4 as possible. The idea seems to have been that by making it harder to move to v6 piecemeal it would make the transition happen faster. If that is the case, then I think that the opposite is the result. By increasing the cost of a slow rollout of v6 it made it more attractive to stick with v4 as long as possible.
I may be completely wrong on this, I will admit. There may be technical reasons why a move to v6 couldn’t have been very smooth. But I’m thinking of NAT here. When NAT came out it was a godsend for many reasons, even though it was a kludge that has a whole new set of problems. It hit where it needed to, by allowing new nodes to be added with very little disruption to existing infrastructure. IPv4 mapping seems to be trying to provide a seamless move to v6 but either it’s not as good as it could be or it has really bad PR.
4. IPv6 readability.
IPv6 addresses are hard to read. There, I said it, you can now come to confiscate my nerd credentials.
Really, though, I understand that v6 addresses are 4 times longer than v4 addresses and there’s no way that they can be as easy to remember and work with. But the 4 octet scheme of v4 is so ingrained in sys admin culture that it seems crazy to depart from it so much with v6. Sure, eventually we’ll all be thinking in v6 subnets and addresses but it’s a big jump. I can’t help but think that the v6 representation came out of academia instead of from the trenches.
5. Subnetting is still important.
When the move to v6 finally comes, don’t give up all of the knowledge you’ve built up around subnetting. Sure, even the smallest v6 address allocation provides for 65536 subnets each with billions and billions of nodes. But that only takes care of part of the issues with subnetting today. It’s still necessary, although easier, to understand where to divide subnets relative to physical routing. It’ll be much easier, but not automatic.
IPv6 is inevitable, so let’s all keep an eye on the bouncing ball. But don’t panic just yet, the tipping point is still a ways away. It would be nice if the move to v6 was as easy as rolling out a new OS (even though that certainly isn’t easy, but it is easier.) But it isn’t, so we sys admins will continue to do what we always have done: Make it all work somehow.