IP allocation practices has been affected by route filtering.
Prefix length cost study "mostly" done
More/Better data
Data for other hardware vendors??
Most of the information needed does not really exist in useful formats
A /18 may work in a 128Mb router.
A /19 may also work in a 128Mb router with up to 7 or 8 views.
No filtering would require many many Mb of memory (700 or more).
Other factors that will influence this are things like path length.
Sean comments that folks often have great ideas, but don't have time to research them. Sean also laments the fact that there are a number of spins that get placed actions based in engineering requirements. Some of the spins are positive for Sprint and sometimes they are not.
Sean also compliments Jeremy on his work, but comments that he is not addressing issues that relate to the dynamic update problems and the cost of processing such updates.
Finally, Sean is also unhappy with the IANA's unwillingness to make public statements concerning the merits of Sprint's actions in routing filtering. He believes that the lack of such public discourse left a large degree of uncertainty that may have resulted in some of the negative spin towards Sprint in both technical and non-technical forums.
Jeremy comments that the registries are not allowed to make decisions on address allocation that allow for operational considerations, but only those that for "ideal" conditions. He also comments that the press is uneducated when it comes to most operational matters and does not really know what is truely important and what is really minor.
Alan Hannan ask Sean what the "sickness" is. Sean says that it is often difficult to correctly educate those that within a company are tasks with interacting with the press and that lack of education often results in inaccurate and inflamatory reporting. Alan also suggested that the analysis be more than just a cost analysis and be a cost/benefit analysis. Jeremy did not want to make this issue more "political" by attempting to measure benefit.
Curtis Villamizer says that within CIDR-D the idea of filtering should be used as a last resort and his main criticism of Sprint has been a lack of willingness to pursue multiple solutions. Sean discounts this. Curtis and Peter are now engaged in the use of the IRR for better aggregation and the discussion is not really going anywhere.
Curtis also suggest that the other types of aggregation is not being considered. Jeremy says that his study does reflect the actual allocation policies. Netaxs has an aggregation tool available to assist in better aggregation.
Concerning the media, things are going to leak, so why don't we establish better documentation to help educate the press and those that interact with them.
Jeremy remarks that working as a group to establish pricing may have anti-trust implication.
David Conrad remarks that APNIC allocates on the basis of their policies, not on the basis of Sprint's filtering. Sprint's filtering has driven folks to the upstream providers. And, he would like to see more providers to implement these same policies. The allocation policies were not created without input of the community.
InterNIC remarks that the Sprint change has brought more reality to their allocation policy.
Curtis remarks that using the IRR will permit the use of aggregates where only more specific routes are being announced. This permits ANS to delete up to 3000 routes. He also does not agree with David Conrad concerning David Conrad's assertion that Sprint was the only provider to do anything to encourage customers to go to their providers for address space.
Peter remarks that multiple paths cause multiple route change announcements everywhere in the default-free Internet. Containing the size of the tables and limiting the number of flaps will make it possible to build bigger networks.
Randy Bush remarks that we have important work to do.