[ppml] Policy Proposal 2005-1: Provider-independent IPv6

[Vince Fuller]

I ask myself "oh why do I waste keystrokes on this?" but then I do it anyway.
If it is true that one definition of insanity is doing the same thing over
and over while expecting the same result, then I should really stop trying to
provide intelligent explainations about scaling issues on NANOG...

> > Supposed that we approve 2005-1.  If things take off, we could then
> > deploy several hundred thousand routes 
> 
> Before the August 11th deadline for the next ARIN
> meeting? I don't think so. 
> 
> > (on the order of the v4 network
> > today), and then we try to decide that we want to reverse ourselves and
> > have several thousand people undo the work that they've done and deploy
> > a different solution? 
> 
> You are completely missing the point. There is an
> ARIN member meeting roughly once every 6 months. New
> policies proposed more than 60 days prior to the meeting
> can be passed by the meeting and will make it through the
> last call, AC and BoT within a couple of more months or so.
> There is no way that kind of growth could happen in such
> a short period of time.

No, you are completely missing the point. Establishing a policy that
defines "provider independent space" sets a precedent that, for all
practical purposes, will not be revokable. Once early adopters are
granted "PI space", the genie will be surely and truly out of the bottle;
those recipients of a valuable commodity will not be easily persuaded to
give it up nor will their service providers cut-off paying customers by
ceasing to route "de-authorized" prefixes. Those assignhments will be
permanent, just as all SRI-NIC-era assignments ("the swamp space") are
still filling the IPv4 routing table more than 10 years after they were
last granted. Future applicants will rightly cry "foul" when told that
they aren't going to be given the same "PI space" that early adopters
were given. 

> But, let's assume, for the sake of argument, that
> these new PI addresses really take off in a big way and
> projections show that it will be in the tens of thousands
> by year end. The BoT does have the power to halt all PI
> applications until the members can reconsider the policy.
> This would be an extremely unusual action but anyone familiar
> with ARIN allocation statistics knows that the numbers of
> PI allocations you are talking about describe an
> EXTREMELY UNUSUAL SITUATION.

Nonsense. Once something is given it is almost impossible to get it back.
And creating a two-tiered society of "early adopters with PI space" vs.
"late adopters who get something less valuable" is going to run into all
sorts of issues.

And the hundreds of thousands of routes that are of concern to those who
think about "big Internet" scaling problems are what we *will* see if a
"PI space" policy is adopted. Not by years end but most certainly if ipv6
is widely adopted in the absence of a scalable multihoming solution (and
no, the existing ipv6 strict topology-based addressing plan doesn't cut
it, either; the protocol MUST be fixed to replace the "address" with a
topology-based locator and a endpoint id that does not need to ever be
renumbered once assigned before ipv6 can be ubiqitously deployed).

> Nobody has proposed that genies be let out of bottles.
> What has been proposed is loosening the belt for a while
> to enable businesses the freedom to deploy IPv6 networks
> the way that they want to. Belts can be progressively loosened
> or tightened as many times as is necessary.
> 
> If IPv6 PI does prove to be a bad move, it is BETTER FOR
> ARIN TO MAKE THE MISTAKE and then correct it. Otherwise we
> risk being the target of restraint of trade lawsuits. It simply
> is not legally prudent to disallow PI allocations in IPv6 and
> it is not legally prudent for the representatives of large
> ISPs to vote against 2005-1 without having consulted their
> legal/regulatory departments.

Ooh, the "restraint of trade boogieman".

I would be rather more concerned over lawsuits regarding unfair and
inconsistant business practices should an RIR establish the precedent of
granting a valuable asset ("PI space") to early adopters and then refusing
to offer equal treatment to later adopters. Or of tring to revoke the use
of an existing "PI space" allocation that is in use for critical business
purposes.

Some organizations were good enough to transition off of pre-RIR IPv4
assignments (anybody remember net 36.0.0.0) because of a sense of community
and of doing good for the collective; that sense of community has long since
vanished from the commercial Internet, so past behavior is most certainly
not a good indicator of future behavior.

Fix the routing and addressing architecture *NOW*; don't try to kludge
around it and hope that it is possible to retrofit a fix later.

	--Vince