On 30-nov-03, at 4:17, Valdis(_dot_)Kletnieks(_at_)vt(_dot_)edu wrote:
The "at current burn rate" assumption is far from safe though...
Oh? Have any better-than-handwaving reasons to suspect the current
allocation rate will change drastically?
I have a slightly better than handwaving indication that the current
statistics don't show the full picture. In the past 4 years we've seen
large scale always-on internet access deployment. However, this doesn't
show up in the address space usage statistics. So addresses must be
coming from other places than just the IANA storage rooms.
Much of the world isn't online yet, but quite frankly,
those areas have severe infrastructure and economic problems to resolve
before they start chewing up a lot of address space (yes, China and
have enough warm bodies to burn out the address space - they don't
monetary units to do so).
It seems there are (or were) 450 million bicycles in China. Think about
it: what's cheaper to mass produce, a 20 kilo steel bicycle with lots
of intricate mechanics, or a simple 1 kilo plastic sub-laptop?
I'm more than happy to accept any realistic projections that point to
a change in the burn rate - if you know of something I've overlooked,
please enlighten us....
I believe there are many factors that aren't accounted for in the
current projections. For instance, some ISPs use non-RIR space they've
held for a long time to give to customers. (See 18.104.22.168/8.) Another
factor is that when people change ISPs, they have to renumber. Since
the policies, and especially the way they are applied, have grown
progressively more strict the past ten years or so, this often means
the organization in question gets fewer addresses from the new ISP than
they used to have at their old ISP. There is more, but let me finish by
observing that it's not much fun anymore to announce a /16 let alone a
/8 over BGP due to all the worm scanning that's going on these days.
The current IPv4 address consumption statistics are like measuring a
pile of books stacked on a big pillow. The stack doesn't seem to grow
that fast, until all the air is out of the pillow and it doesn't
compress any further. Then the pile starts growing higher much faster.
So while I completely agree there is no reason to worry about IPv4
address depletion the next five years or so, I wouldn't count on
significant numbers of IPv4 addresses being available in 10 or 20