In message <Pine.LNX.4.30.0305291823140.17378-100000(a)boggle.ihug.co.nz>nz>, Simon L
On Thu, 29 May 2003, Joe Abley wrote:
So, only 14% of the available IPv4 address space
had been assigned to
RIRs at the time of that presentation.
Which is fairly meaningless since half those 202 are instead delegated
to Large Organisations, Various Registries etc.
For sure. But if push comes to shove IANA can talk quietly (perhaps in
the "talk quietly and carry a big stick" sense) to some of those Large
Organisations and see if they couldn't perhaps manage with somewhat less
than 2**24 addresses.
As the cost of putting a full IP stack and connection
drops to only a
couple of dollars or less all sorts of random things will start having
them. And before you say NAT remember that the power company will want to
directly access your meter and hotwater cylinder, your car company the
car, your whiteware company the fridge etc. Did I mention the games
For many of those it could easily enough be done with NAT. It seems to
me that "phoning home" isn't, inheriently, any more problematic than being
able to interogate the devices remotely for the vendor. And it's probably
somewhat preferable for the customer too. (Even if everything was globally
routable, would you want Joe Random Company connecting in to arbitrary
devices your network? This isn't your grandmother's Internet any longer.
Sure you can use a firewall. You've just moved the problem to being
one of ensuring a million or so firewalls are properly maintained. If I
were a vendor, I'd take "phoning home".)
When stuff like that comes along we really want them to
be deployed on a
nice ipv6 internet rather than have to suddenly switch cause demand
increases by a factor of 10 overnight.
It would be nice to have $1,000,000 tomorrow too. I don't think either
of them will happen.
Even at the current ease of use the cost/benefit ratio for converting to
IPv6 is too low for (m)any(one) to put effort into doing so. It'll only
be when there's a whole lot more benefit (eg, lots of IPv6-only
services, or stuff which cannot be done with IPv4+NAT) or a lot less
effort (works out of the box without even thinking about it, and has for
years) that'll actually happen.
Especially since we've had the better part of 10 years for everyone to
get pretty good at designing IPv4+NAT compatible protocols and networks,
thus greatly mitigating the perceived benefit of switching.
 Although should someone wish to give me $1,000,000 tomorrow, I'd be
willing to look suitably surprised on camera.