...

Would someone be so kind as to email me a list of the Exchanges that matter? I.E., which ones qualify when calculating your connections for peering?

Is it just the 4 NAP's and MAE-East and MAE-West?

I thought of answering this but I decide that followups to other folks replies would probably be more useful to the gallery. I was right:

Ameritech NAP Atlanta NAP :-) MAE-East MAE-West PACBell NAP Sprint NAP

In terms of meeting peering requirements, the above list is sufficient except that the ":-)" has to be read as "just kidding". Unfortunately for my ulcer, more was said:

CIX may be worth connecting to as a sales thing, and PAIX is not worth connecting to yet because there are so many NAPs in that area.

Connecting to CIX won't help your sales. It does help your connectivity if you aren't otherwise able to buy T3 lines to everywhere in the universe, and it's a fine backup for folks who _can_ afford T3 lines to all of known space. PAIX is the best NAP-like object in the Bay Area, in my biased view (I'm a consultant to Digital so the bias is strong). They have better facilities than MAE-W and they aren't subject to ATM's cell tax and PUC vagueries the way Pac Bell's is. The only thing they don't have is a lot of people to peer with, which is a good reason _but_the_only_reason_ why they are not in first place on the west coast. The higher quality of the facilities and remote hands at PAIX ought to lead most newcomers to the Bay Area to locate their POP in Palo Alto and run a T3 line to MAE-W or PB-NAP or both. The remote hands people at PAIX are not knuckle dragging frame techs, they are senior-sysadmin-quality technical people who you would be lucky to be able to hire full time if they were available. This makes a huge difference when you want to know WHICH red light is blinking. On the other hand the original question was about what you need to connect to in order to meet Sprint's or AGIS' peering requirements, and the original answer (a) was correct and (b) did not list DEC PAIX. Therefore I'm really not trying to change the answer, I'm answering an entirely different question.

...

Also, do any of you have any comments on whether peering standards might be relaxed if you are setting up a statewide educational network?

Hmm, possible, but don't count on it. If I were you, I would get connected to the NAPs ASAP because it will get harder and harder to peer with the big guys as times moves on.

The big guys are already yanking existing peering sessions down when they change their requirements. Getting in early has done nobody any good so far. We're seeing bifurcation into little guys and big guys. It remains to be seen which end of the strata will have the larger total number of endpoints. If it's the big guys, they will squeeze the little guys no matter how many of the little guys band together. If the little guys own more endpoints then some kind of collective bargaining will be possible. My hope rests on this possibility but please don't bet your money on it.

On Nov 29, 1996, Nathan Stratton wrote:

Yes, true, it would be great of MAE-West of PACBell just moved to PAIX, but I don't think it is worth adding a view when you have established NAPs in the area. I learned this when I when I was building Atlanta-NAP I wanted to build a place better then PAIX (and they did the best job so far), but I found out that people don't care. If I build a NAP that was 100 times nicer then MAE-East (and that would not be hard at all) people would not just move.

Asking people to relocate their high-bandwidth peering point connections is not likely to be successful unless you give them a cheap, easy way to do it. Using the MAE-east/Atlanta-NAP example, why would people pay thousands more dollars per month to backhaul traffic to Atlanta when they can exchange it locally? It just doesn't make sense, and really defeats the purpose of having a local exchange point. Sure, the number of hops would be the same, but if a packet has to move between two DC-area providers, it makes little sense for it to travel down the east coast and back up again. But the biggest reason that people will not move their peering point connections from one place to another is that it will break things. Also, it will be a huge pain in the ass. If things are functioning reasonably well, there is little point in moving everything around. Alec -- +------------------------------------+--------------------------------------+ |Alec Peterson - chuckie@panix.com | Panix Public Access Internet and UNIX| |Network Administrator/Architect | New York City, NY | +------------------------------------+--------------------------------------+

Could you expand on this a bit please? For instance, how do you define "endpoint"?

I misspoke. I shouldn't put this in TCP terms, it's just that TCP is the economic driver right now. But what I mean is "host", whether a ppp client or a web server or a dns server or whatever. I'd like it if there were some way to weight endpoints with some kind of "desirability factor" so that we could more meaningfully charge each other for access to our nets. Right now it's assumed by most peering contract authors that all endpoints are of equal economic value to other endpoints, and all endpoints derive similar subjective value from having access to the net. None of that is true but we have no way to measure the truth in this case. Yet. It's more of an anthopology problem than a technology one.

On Fri, 29 Nov 1996, Derek Elder wrote: |} So is there a better solution for someone that doesn't actually want a |} POP there now? If PAIX is the smartest connection facilities wise, |} that is where we want to be. Are 2 DS-3 connections per router allowed |} at PAIX? That would be the best solution -- cohab a router at PAIX, |} DS-3 back to our core and DS-3 up to Pacbell. |} |} What am I missing? Just that you'll have to eat extra circuit costs. While PAIX is a nicer, newer and cleaner facility than what MFS has in San Jose for MAE-West; I don't recall too many failures or complaints about the quality of that facility. Ask BBN where their SJ POP is located. Since I don't think this list has turned into 'network-engineering' I'll stay away from the P*B NAP vs. MAE-West point. ;-) -jh-