Policy Proposal 2002-3

[Shane Kerr]

> > Policy Proposal 2002-3: Micro-Assignments for Multihomed Networks

On 2002-09-23 14:03:36 -0600, Trevor Paquette wrote:
> 
> I would also add that the organization must continue to be
> multi-homed. This should be able to be verified via BGP
> advertising.
> 
> When an entity is no longer multi-homed for 1 month (anyone
> else suggest another timeframe?), ARIN should reclaim that IP
> space and inform the original organization and the single BGP
> advertising entity that they are to stop advertising it.
> 
> My concern is that there needs to be a built in reclaimation
> process as well as a provisioning process.
> 
> How might this be monitored and who is going to monitor it?

There was a fairly detailed discussion on one of the RIPE mailing
lists about reclaiming AS numbers that were no longer multihomed.
The thing is, how do you determine if a route is multihomed or
not?  You can't look at the BGP table, because BGP doesn't
forward all paths - just the "best" one.

You might be able to reclaim some resources by allowing for an
imperfect solution.  That is, if ARIN, and perhaps some friendly
ISP's willing to give ARIN access to a BGP dump, doesn't see a
resource being multihomed, then it could ask for the user to give
the resource back.  If the user declares that it really is
multihomed, even though it does not appear in the BGP feed, they
get to keep it.

While this would allow some users to continue to hold unused
resources, it would recover some from either defunct
organisations, or merely forgetful and/or lazy parties.

The RIPE folks decided this was a bogus idea.  Personally, I
think the RIPE community was too hasty in dropping the idea.
Encouraging users to return some unused resources in a relatively
painless way seems like it could be worthwhile.

> Nothing like burning up the remaining IPv4 space to speed up
> the roll-out of IPv6......

To be fair, all the RIR's combined have only allocated something
like 11% of the IP space over a 10 year period.  How long do we
want the IPv4 space to last, anyway?

-- 
Shane
Speaking only for myself