Interesting People mailing list archives
must read really FCC NPRM ban on Paid Peering harms new innovators
From: Dave Farber <dave () farber net>
Date: Thu, 12 Nov 2009 17:08:12 -0500
Begin forwarded message:
From: Stan Hanks <stan () colventures com> Date: November 12, 2009 4:07:33 PM EST To: dave () farber netSubject: RE: [IP] Re: FCC NPRM ban on Paid Peering harms new innovators
Wow... This all brings back a ton of stuff... So, in the early 90s, I was involved in the creation of the MAE Eastpeering point. Initially, it was a "Metropolitan Area Ethernet" to solve Steve Wolfe's problem of not getting Marty Schofstall and Rick Adams toagree to play nice and meet the NSFnet and CSnet in a convenient location, but as we all know, it exploded into something much larger.The MFS MAE model was simple "we don't charge you for moving packets, we charge you for ports, and you work out the details on exchanging packetswith whomever you want to exchange packets with on your own". In practice, the model that DJF noted was the rule of the day -- to avoid having to address the "settlement" issue, everyone agreed to pretend like traffic exchange would be in parity and thus no settlement wasneeded. That worked *great* when there were only a handful of networks, and when MOST of the traffic was "classic ARPAnet style traffic" -- FTP,email, TELNET, etc.The advent of the "world wide web" changed all that. Suddenly two thingshappened -- first, there were a HUGE number of entrants to thenetworking game, and second, traffic flows became inherently unbalanced.At the last point where I had personal knowledge, which was probably '94or '95, there were close to 300 parties connected to MAE East. I understand that it went up from there. That's an INSANE number of network peering relationships to manage!The very nature of "web pages" also was a big deal. Whereas with email,it's reasonable to assume I send you one, you send me one, we'reprobably writing about the same amount, on average it all balances, web pages just don't work that way. Send a very small number of bytes askingfor a page to be served, get a potentially unbounded number of bytesback in the form of text, image, video, what have you. Instant nightmarefor people trying to manage traffic on networks. When I was putting together what became Enron Broadband, one of the early directions we looked at was becoming a major "Tier One" ISP backbone. At the time, "Free Peering" was the rule of the day, but was already dying. If you didn't have peering established, it was verydifficult to get. My engineering team put in a lot of time and I pulled a lot of personal relationship strings, but it finally became clear thatif we wanted to do this, and have free peering, the only way to get there was to acquire players who had the peering that we wanted. That was in '97. By 1999, it was pretty much all over. Level 3 was talking in terms of "suck" and "blow" -- are you pulling traffic out, or are you sending traffic? And if you were sending them more traffic than they sent you, you had to pay. Period. You see that replayed countless times since, the most recent being in the dispute with Cogent. The economic argument is interesting. If you're a content provider, a "net source" of traffic, you argue that the end user subscribers areonly customers of their network provider in order to gain access to theinformation provided by the content providers, and thus the network providers should be glad that the info is there because otherwise they wouldn't have any customers. Of course, the network providers take the obverse position...My take is that from a amount of work and investment of capital, it's a LOT harder and more expensive to provide an operating network than it is to provide a web server farm, and thus my bias is towards the guys withcontent paying the guys with pipes to get it delivered. But the truth is, it doesn't matter. If you want to do business, youhave to accept that it's just a matter of economics, and you either have to figure out how to make the economics of paid transit work for you, oryou have to figure out how to scale large enough that you can build enough of your own infrastructure to pose a credible threat to the network provider establishment that they're willing to modify their terms for your benefit. Stan
------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com
Current thread:
- must read really FCC NPRM ban on Paid Peering harms new innovators Dave Farber (Nov 12)
- <Possible follow-ups>
- must read really FCC NPRM ban on Paid Peering harms new innovators Dave Farber (Nov 12)