On 7 dec 2007, at 10:14, Rémi Denis-Courmont wrote:
However, although I agree it's true in the general case, I think for non-trivial servers/services, this isn't true: if there is a good reason to make changes on the IPv4-only side, it may be possible to make them.
That would depend on your definition of "non-trivial". But I am afraid Jari is right - host still lacking an IPv6 stack are either not upgradeable, or notwilling to be upgraded.
The case I'm thinking about is a service that is somewhere up there in the top 10000 destinations on the web. That's big enough to make deploying IPv6 non-trivial (especially now that Vista boxes enable 6to4 when they have a public address even if protocol 41 is filtered so they can't reach destinations with a AAAA record) but the operators of such a service would presumably be willing to install reasonable updates if that buys them something useful.
I think we really should start considering a transition phase where the clients may run IPv6 but the servers largely don't. (I know, home users can also run servers, peer-to-peer, all that good crap.)
As someone pointed out already, modifying even existing IPv6-capable host is pretty difficult: it would be 15 years or so before the Windows more recentthan Vista would have a fair share of the market, if it ever happens.
Hm, pre-XP windows is quickly disappearing and both XP and Vista should be considered updatable if Microsoft chooses to do so. So in another few years the main class of unupgradable IPv4 hosts will be special purpose devices rather than computers running a general purpose operating system. But obviously updatable in principle doesn't mean it will happen on 100% of all boxes.
Also worth considering is the self-realizing prophecy factor - if >80% of IETFers think that a given IETF technology (say modifying IPv4 stacks to support IPv6 transition) won't happen, how likely is it to be successfulanyway? :(
After wednesday's plenary? Highly likely. :-)Note that I wasn't talking about upgrading ALL IPv4 hosts, I was just saying that SOME classes of IPv4 hosts can very likely be upgraded.