Mea Culpa (was Re: LET'S JUST GO AROUND)
I'd like to thank Scott Bradner for injecting some actual hard info
into the discussion (via a private e-mail to me). I'm sure that
Scott will chastise me appropriately if I've mis-paraphrased what
he told me, and everybody else will flame me if I added 2 and 2 and got 5. ;)
Scott pointed out to me that at least at first, the ARIN registry will
operate with exactly the same policies as the already existing registries,
and that any changes in said policies would have to be thrashed out and
agreeed to by all the registries, on a *different* already existing list.
Taking that info, and poking around www.ra.net for RADB information, I
eventually stumbled onto a document RIPE-140, which stated the policy
for the European side of the puddle. Section 3.4.2 of said document
basically answered the question re: multihoming, by saying that even if
a request was for provider-independent addresses, the registry should
trim it down to the smallest CIDR-sized/aligned block that would fulfill
the address regardless of routing.
My conclusions after reading the additional documents are:
1) The *current* policy doesn't cut any slack for people wishing to multihome.
(I'm assuming that if the RIPE document doesn't, that the other registries
and IANA don't either, via implied transitivity by having similar policies).
2) ARIN won't either, unless there's a global change in policy.
3) The issue of too-small companies trying to multihome is mostly a
moot point, as apparently one of the following has happened for every
case of what I was concerned about:
a) They were big enough to get a globally routable address
b) They decided to go with redundant links to a large ISP such as NetAXS.
c) They managed to bribe those long-haul people with filters to pass their
d) They appealed directly to IANA and got an allocation anyhow
e) They were picked up by the guys in the black helicopters. ;)
In any case, we've not seen any actual lawsuits about this policy, so
I guess ARIN adopting it won't change things any. And I *know* that we
heard about at least the first several lawsuits against NSI re: their
domain name policies...
I personally remember multihoming being a contentious issue when CIDR
was first designed and deployed, and apparently the question of exactly
what a registry (or anybody else, for that matter) should do regarding
multihoming and its impacts on CIDR is still an open research question...
TIme for me to go off, eat some crow (who's got the tabasco sauce? ;),
ponder the RADB databases, and shut up on this particular issue until
I have some hard numbers regarding multihoming and CIDR ;)
Computer Systems Engineer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 284 bytes
Desc: not available