> On Nov 22, 2022, at 9:09 AM, John Curran <jcur...@istaff.org> wrote: > ... > Interoperability isn’t insurmountable, but does take some investment of > effort. One can imagine any number of techniques (e.g. flag day after which > “production devices” on the Internet must support 240/4, or DNS resolver > hacks that fail to return “A” records with 240/4 addresses unless a flag is > set that says “we’re in the 240/4 routable Internet” [ick], etc., etc.) It > doesn’t seem particularly hard to come up with some approaches to solve the > interoperability problem, but completely ignoring it is not an answer (and > makes it rather difficult to take your proposal seriously…)
Joe - By the way, you shouldn’t feel particularly bad about skipping out on the interoperability requirement – anything involving interworking with the installed Internet is hard, and this is the same lesson that the IPv6 folks found out the hard way… I will confess that I was a member of the IETF's IPng Directorate and thus inherently complicit in that particular fiasco – With IPv6, the first answer to interoperability was “let’s do tunnels between IPng islands”; i.e. helpful for lab environments but useless otherwise. We then declared that transition was a problem “to be solved later” but that shouldn’t get in the way of the declaration of IPng as the new IPv6. Finally, after failing to solve the problem, we reverted to “ships in the night”; i.e. IPv4 and IPv6 running in parallel on the same infrastructure – it works, but defeats the entire idea of IPv6 as a functional substitute for IPv4 for connecting new customers and infrastructure to the existing IPv4-based Internet (Luckily, the service provider industry that was growing most rapidly realized that they really needed IPv6, and they really needed transition solutions that allowed IPv6 to interoperate for IPv4 for new connections, and so we eventually saw real solutions such as 464xlat, ds-lite, etc.) Maintaining interoperability with the existing base is hard - far harder than just “updating the standard” - but is absolutely essential if you want viable reuse of 240/4. Of course, it does raise the question of whether the total effort will be worth the purported gain, but that really can’t be assessed until there's some specification of the proposed solution for interoperability with the existing deployed devices that don’t know about the 240/4 change. Thanks, /John p.s. Disclaimer(s): my views alone. Warning: may cause dizziness, headaches, or nausea.
