Because of the fact that the NAPs are a status symbol. They are a marketing tool that private peerings are not quite yet at. The mass market is just starting to realize the NAPs existance and the whole evolution of the NSF backbone into this distributed IXPs. Therefore, on that understanding, AGIS is pretty damn elite. OR... Are they going to a bunch of IXPs because none of the TRUE big boys won't peer with them over private exchanges because they can't justify the traffic levels that the TRUE big boys have for private exchange peerings.. the only one I know a # figure on is Sprint's, which is 90megabits sustain> Many different models here.. you have PSI at mae-east, mae-west, UUNET and BBN at sprint, mae-east, mae-west, ANS at sprint, pb,mae-e,mae-w MCI and Sprint at sprint,pb,mae-east,mae-w,aads. Who's right? Who's wrong? Who cares? Rob

Here is the question that occurred to me.

If we set a requirement to be at 5 NAPs, and we don't peer with anyone who isn't at 5 NAPs, and we only peer with like 4 networks that qualify, aren't we essentially talking about using GigaSwitch or ATM switches as private interconnects? [barring the defaulting issue on FDDI]

If whole point is to exclude networks due to a number of technical reasons why go to 5 or 20 NAPs when private connects would serve the same purpose? Or is it some kind of bragging thing where a network can say "We went to the time and expense of engineering connections to 5 NAPs and now no one qualifies for peering with us." Wouldn't the obvious question be, "Why did you bother then?" For several organizations it isn't the money that is really a question with multiple NAPs, but the marginal value of the next NAP after you are already at 3 or 4 whatever is considered acceptable/comfortable. [Economic theory, sorry]

Anyone agree?

-Deepak.