[arin-ppml] Draft Policy ARIN-2024-5: Rewrite of NRPM Section 4.4 Micro-Allocation

[Bill Woodcock]

> On May 24, 2024, at 03:54, Martin Hannigan <hannigan at gmail.com> wrote:
> ...The other is IX preparing and certifying peers, getting resources but then never deploying a switch fabric. Wanted to have a good revocation trigger. Likely to be used rarely if ever, but for thoroughness. Neither are corner cases. 

Indeed.  There are about 800 IXPs, currently, and there have been about 400 others that are now defunct.  So, thus far, about a 2/3 success rate over slightly more than thirty years.  The mean-time-to-failure is, as Marty points out, a front-loaded, long-tailed curve…  Most IXPs that fail fail very quickly, never _really_ getting off the ground.  Others fail less predictably, at some point down the road, when conditions or politics or regulation change.

We haven’t done a study of IX switch-fabric prefix reclamation…  That would be a good short undergrad thesis project, and would be useful.  I’ll suggest it to a few faculty and see if there are any bites.  The big shift that would probably be visible in an analysis would be the renumbering out of NET-198-32-0, which is itself a valuable lesson about the cost of poorly-defined policy.  

My casual observation is that often IXP switch fabric subnets get allocated to an IXP effort in a metro area; that IXP effort comes to nought; some time later, another unrelated IXP effort happens in the same metro, and uses the already-allocated space.  Which seems ok in practice, but a little ad-hoc and probably not something that should be depended upon.

                                -Bill

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: Message signed with OpenPGP
URL: <https://lists.arin.net/pipermail/arin-ppml/attachments/20240524/bc9f6874/attachment.sig>