At 9:48 AM 10/27/94, Martin L. Schoffstall wrote:

Merit/NSFNet already tackled these issues in an insufficent and unopen manner.

MarkFedor/ColeLibby from PSI said there was a "quiet" admission that the old methodology was already "approved" for the SPRINT NAP.

Marty

Since it is not possible or desirable for NAP providers to collect statistics on IP source/destination pairs, protocol types, or per-NSP datagram counts, it seems that a different mechanism would be needed to do engineering analyses to manage Internet performance and growth. The NSFNET mechanism which used a device at each regional DMZ (nnstat or equivalent) seemed like a good mechanism to do that, especially given the research/education focus of the regionals. I suggest that it would be very difficult to do this sort of monitoring on commercial networks. Why not continue the NSFNET sponsored and approved method of placing collectors at NSF-supported regional network DMZs even after the NSFNET transition, so that we have at least a sampling of Internet traffic measurements. Mark