[ppml] Draft 2 of proposal for ip assignment with sponsorship

[Alec H. Peterson]

--On Thursday, February 27, 2003 3:14 PM -0800 william at elan.net wrote:

> We're not in ipv6 age yet. In ipp4 renumbering, no matter how well the
> network is setup, is difficult - especially when dealing with large
> enough  network. You have to change setup in most several routers
> (include firewall, switches as well as changing filters, change setup in
> many computers  (even dhcp will not completely solve things), change dns
> (both nameserver  ip and actualdns zones). So let me respectively
> disagree with you about  how easy renumbering is.

Fair enough.

> I agree with you. And considering widespread difference in opinion on
> this  micro-assignment issue we may not be able to get everybody to
> support any policy proposal so a compromise proposal that has support of
> majority but  may not go far enough for some but too far for others maybe
> the answer. This is a tipical situation with the how laws are made.

We'll never get 100% concensus, but we need to be closer to where we are 
now.

>
> And on specific points you have above:
>  honesty: Proposal has to provide for good way to catch people who are
>           lying and to minimize the abuse. Many believe that relying only
>           on ARIN staff to catch those who lie does not work (you said it
>           yourself!). ARIN becoming net-police also is not an answer to
>           stop possible abuse.

True.  And changing policies just because people lie is also a poor idea.

>  renumbering: I think I already said enough on this above.
>  routing table: Multihoming organizations are already announcing part
>           of their upstream's block as separate BGP announcement. Having
>           them announce their own block would not change size of the
>           routing table, but good provisions must be put to check that
>           those requesting micro-assignments are indeed multihomed.
>           Relying on them just having an ASN without futher verification
>           may not be enough.

But there is one key point that you have missed (and many people have 
missed).  While the number of prefixes may not change, the structure of the 
table will change.  Right now, a /24 out of one of UUnet's /14s is part of 
a larger aggregate.  If this UUnet customer has his own /24 this is not 
part of a larger aggregate.

What is the impact of this?  Well, with today's routing table and routers 
nothing.  However, history has prooven that it is sometimes necessary to 
not accept all announcements, and the easiest way to deal with this is to 
filter on RIR allocation boundaries.  If we move forward with having ARIN 
allocate /24s then we are tying the hands of the backbones that we all 
depend on.  Making these microallocations out of a separate block would 
help mitigate the issue, but there would still be far more prefixes out 
there that are not part of smaller aggregates, which is the fundamental 
issue here as I see it.

As I have said in the past, there are so many new mistakes we can make, why 
must we insist on making the same ones again?

Alec

--
Alec H. Peterson -- ahp at hilander.com
Chief Technology Officer
Catbird Networks, http://www.catbird.com