On Sat, 20 Nov 2021 at 16:31, Leo Vegoda <l...@vegoda.org> wrote: > On Sat, Nov 20, 2021 at 8:06 AM Matthew Walster <matt...@walster.org> > wrote: > > On Thu, 18 Nov 2021 at 15:04, Leo Vegoda <l...@vegoda.org> wrote: > >> > >> On Thu, Nov 18, 2021 at 6:05 AM Tim Chown <tim.ch...@jisc.ac.uk> wrote: > >> > Personally, I’d say just get on with IPv6. > >> This gets my vote. > > > > So... This is tricky. IPv4 isn't going away for residential and most SME > connectivity -- it's just too difficult to live without, and too entrenched > in the psyche of the IT tech support PFY. > > [continues well argued point] >
Awww, thank you, kind Sir ;) > Yes, but there's a distinct difference between not rushing to remove > RFC 1918 address space from most home and office networks and putting > lots of effort into trying to make some other, previously reserved, > IPv4 space reliable enough that it can be uniquely registered to > various network operators. > Oh absolutely. For the record, I'm against any of these proposals, but the only one with any merit to me is 240/4... And that was a debate raging on NANOG back in 2007 when I first joined the internet community as a network engineer: https://archive.nanog.org/mailinglist/mailarchives/old_archive/2007-10/msg00451.html However, I remember Canonical trying back in 2015 and not getting anywhere with it: https://ubuntu.com/blog/introducing-the-fan-simpler-container-networking -- it would be nice to have 240/4 as private address space, sure, but it's literally not worth the magnitude of effort and solution-lockin compared to the alternatives. I'm just saying that screaming "IPv6 is the answer, deploy it" is actually forgetting that it doesn't actually reduce IPv4 usage in and of itself, which is the goal. What IPv6 needs is feature parity with IPv4, not from a technical perspective, but from an operational one... and I can't think of how to easily solve that without a well-known, short, and memorable prefix that is NAT66 to the outside address by default. Sure, you can have a preferred prefix announced via RA as well, similar to the mechanism ULA uses but decidedly "not unique", as being able to fall back to that well-known addressing would make a whole lot of sense for the operational burden. I'll be quiet now, I've spoken enough :P M
