v6 subnet size for DSL & leased line customers

I'd also be interested to know what you'd *want* if you were asked how
you'd like to structure IPv6 addressing, if you didn't have any history
of having to be conservative with IPv4 addressing. IOW, imagine IPv4
didn't exist, and therefore your thinking about IPv6 isn't influenced
by your history with IPv4.

Well since it looks like it takes about 20-30 years to get a new version
of the IP protocol deployed we have to look way ahead.

Now I think there is a chance that full nanotech could deploy in the next
20 years so the protocol should probably be designed with that possibility
in mind.

Now according to an article on Utility Fog [1] one idea is that most of
the household objects around us could be replaced with small nanotech
robot. Each might weigh 20 micrograms which means 50,000 per gram or 50
million per kilogram. Non moving CPUs would probably be smaller.

So my house may have a couple of tonnes of them scattered around in
thousands of objects (chairs, screens, door handles, fly screens, sensors)
with between a few hundred to a few billion bots in each.

Yeah sure it's science fiction now but it's fairly possible that it could
be the situation in say 2030 and IPv6 is probably good enough to handle
it. If we'd let you design a protocol that didn't support billions of SOHO
addresses then around 2020 we'd be madly deploying IPv7.

However in the shorter term nobody has billions of IPs and most people
don't have thousands of networks.

My understanding is that the main idea with the /48 is that everybody
smaller than a provider, government or a Fortune 500 company will just get
one and no further paperwork will be required. Dropping it to a /56 means
that a certain percentage of your customers are going to have to
negotiate, fill out paperwork and pay extra for their /48
which is going to add costs all around.

[1] http://www.kurzweilai.net/meme/frame.html?main=/articles/art0220.html

logic chains which begin with

Now I think there is a chance that

may not be the best way to do engineering. there is a 'chance that'
just about anything.


Sure, the Sun could explode tomorrow and all these IPv6 people will have
wasted their lives.

However the scenario is a common one and the timetable is well within the
time period when IPv6 will be the main network technology (say 2015 - 2035+)
so it should have been taken into account and judging on the fact that
IPv6 *does* support billions of nodes and thousands of networks to every
end site I guess it probably was.

Making engineering decisions on the basis of "there is no chance" that
is risky too, especially looking 40+ years into the future.

simon, there are a million chances. and we are notoriously bad at
predicting any of them more than a year or so out.


In a message written on Sat, Dec 22, 2007 at 12:29:54PM +0900, Randy Bush wrote:

simon, there are a million chances. and we are notoriously bad at
predicting any of them more than a year or so out.

Perhaps the take-away is that we shouldn't try to design protocols
that last for 30-100 years, but rather design frameworks that make
deploying updates easier?


In general, you're right. But we have ~60 years of experience teaching
us that *every* successful computer architecture runs out of address
space. I see no reason to think that today's models for home address
space will be the exception (unless, of course, IPv6 is unsuccessful,
but in that case it doesn't matter much what we do for allocation
policy unless our actions are sufficiently stupid to cause the failure).

    --Steve Bellovin, http://www.cs.columbia.edu/~smb