The latest

[Jim Fleming]

On Friday, June 27, 1997 2:29 PM, Stephen Sprunk[SMTP:spsprunk at paranet.com] wrote:
@ At 08:58 27-06-97 -0500, you wrote:
@ >On Friday, June 27, 1997 8:40 AM, Stephen
@ Sprunk[SMTP:spsprunk at paranet.com] wrote:
@ ...
@ >@ If you'll remember, I argued for atomic netblocks in NANOG a while back and
@ >@ received no support, even from you, who claim to support the idea.
@ ...
@ >Again...I have suggested a very simple plan...
@ >
@ >3,000 new /18 allocations for ISPs.
@ >An easy application form with NO "subjective" decisions.
@ >I can dig out the details if you missed the original postings.
@ 
@ What is your reasoning for the allocations to be /18s, and for there to be
@ exactly 3,000 of them?  I'd have suggested 4096x /20s (exactly 1x /8),
@ except for the problem with Sprint (and others) filtering things longer
@ than /19.
@ 

3,000 was a number someone threw out on a list
in a similar discussion. I think that it was based on some
estimate of the number of U.S. ISPs that would apply
the first day.

The short form approach would require that the ISP
prove (via signed affidavits) that they have contracted
for two distinct "upstream" providers each with at
least a T1 connection.

As for reclamation, maybe these affidavits would have
to be "renewed" each year (like insurance certificates)
and if they are not provided then the IP addresses are
reclaimed in some period of time (30, 60, or 90) days.

Another option would be to use the DNS system to
set up some dummy zones which contain delegations
to companies that then record their connectivity in
TXT records. With that approach, software could be
used to detect when connections are pulled or changed.

In general, if the allocations are large then the
reclaimed blocks will be large. If the blocks are the
same size then the management tasks are reduced.
I think a /20 is a little small. A /19 is OK, but a /18
should hold many ISPs for some time. If they go
belly up, then a /18 is returned.

--
Jim Fleming
Unir Corporation