howland@Priss.com (Curt Howland) writes:

...

The 'Net gives some measure of universal connectivity because it was driven at its root by engineering. Who on NANOG, with any real world experience in networking, denies that maximum open peering benefits everyone?

Well, I dunno if I'm qualified by that measure. At the time I built networks, I'd built at least one of the largest anywhere. But they were all small compared to today's second tiers, and I'm not building networks any more.

:-)

But I deny your assertion. Maximum open peering benefits some people more than others -- specifically the ones who get to charge the most money yet who pay the least in infrastructure upkeep. A web hosting company doing shortest-exit (no matter how many peering points they were at or how much private peering they had) would be an example.

There are two key points to be made here in the general case, and one of a more specific nature. 1. As you stated, benefits some more than others. That statement alone agrees with Curt's assertion that it benefits everyone. You merely state that said benefit is not equally distributed. I'll give you that, to some extent, but not completely. 2. A hosting company doing shortest-exit doesn't necessarily receive more benefit from that peering that the people they peer with. Afterall, the majority of the traffic a web hosting company delivers to any other provider is traffic that other providers customers requested from the web hosting company. As such, it really should be, in most cases, up to the requestor to pay the freight, don't you think? Afterall, when you order something over the phone, they don't usually give you free shipping, and if they do, guaranteed, it's part of the price of the product one way or the other. Same basic scenario here. The more specific point: What about a hosting provider that is willing to honor all MEDs provided by the other provider? That changes the model pretty substantially, doesn't it? If the other provider puts out good MEDs, and the hosting provider honors them, then the other provider is arguably getting the greatest benefit out of the relationship. However, in general, most of the credible hosting companies I have talked to are willing to honor MEDs in peering relationships and are willing to peer at the least on an East+West basis, and often, more geographically specific than that.

Remember as you puzzle your way through this issue that peering is only mutually beneficial if the number of bits (not packets) sent by each side is in the same order of magnitude. If the O(mag)'s differ then the costs/benefits are one-sided, and the side who is underwriting wide area transportation costs for people who aren't paying it money is going to get bent about it.

Again, I disagree. If provider A's customers want data from provider B's customers' servers, then peering is mutually beneficial. Maybe not equally beneficial, but definitely mutually beneficial. The assumption in that statement above is that the transmitter should always shoulder the cost. That's not, necessarily, IMHO, the correct model. I think that the requestor should shoulder some portion, and the transmitter should shoulder some cost. I'm not completely sure what those ratios are. However, if you change the view to one more like mail-order goods, then provider A is collecting money from his customers to cover the freight for whatever goods they request from the customers of provider B and any others. If provider Bs customers feel that they want to subsidise provider A's customers reaching their content, then that's a different issue. Let's face it, there's lots of different ways to look at the costing model on this one and who should shoulder the cost is not really clear to me.

Anyone with a strong enough constitution to check the archives on this matter will find that Sean and I had a raging battle here about this very topic back in 1993 or so. What interests me on this particular night is my memory of asking Vince "so what about gatekeeper.dec.com? why should i have to pay to transmit the FTP archives?" Vince didn't answer because the obvious answer ("gatekeeper should be charging money so it can cover its costs and the costs of the folks who carry those bits") was one I was not at that time ready to hear or understand.

Although that may be the obvious answer to you, Paul, I'm not sure it is the correct answer. Afterall, the people downloading from gatekeeper could similarly cover the same transit costs if gatekeeper was left out of the situation, and instead, the providers with which they already have a business relationship were simply to bill a more appropriate amount for the service they provided to those same customers. Admittedly, it may be easier to convince the gatekeepers of the world than the bulk of the customers on the net, but the same people end up paying one way or the other. It's just a question of how complicated you want to make the accounting in between and how many bills they have to pay from how many sources.

-- Paul Vixie La Honda, CA "Many NANOG members have been around <paul@vix.com> longer than most." --Jim Fleming pacbell!vixie!paul (An H.323 GateKeeper for the IPv8 Network)

Owen