Tuesday, April 22, 2008

The Metropolitan Internet

I am a big admirer of Bob Frankston, even though sometimes I find it hard to understand him.
But I guess this is very interesting:

There is a basic fallacy – the idea that transit necessarily costs a lot of money. And, of course, we confuse cost with price. We are stuck in the idea that the Internet is some distant land that’s hard to get to. So the carriers make decisions that exacerbate the problem. The situation is akin to making it too expensive to put in a path from your house to your garage because you burden it with the cost of a trip around the world.

I don’t care what it costs to get bits from here to Timbuktu – I care very much what it costs to get from my house to my local school. Comcast and the other carriers chose not to peer locally and then tell me that it’s very very very expensive to haul bits from me to my neighbor because I just might want to watch a webcam in Tonga over a satellite link.

If I pay $1k for a fiber from my house to the center of town and so does my neighbor then can we amortize it across many bits even if we stream gigabits each second. Or I can put some inexpensive DSL modems on a copper wire and get megabits. It’s even more extreme for connectivity within my house but at least there no carrier is going to force me to pay for those bits though that was very much what they tried to do in the 1990’s when they tried to argue that connecting more than one computer to a cable modem was stealing from them.

They configure networks as if the Internet were far far away they come up with arbitrary peering prices. That doesn’t meant that it’s intrinsically expensive – it means they’ve managed to make it expensive. What happens if they light up all the fiber they have and follow Moore’s law? We get the fiber bubble bursting again and they go belly up. But that’s not because it’s expensive – it’s because selling transit is a business that does worse the more capacity it delivers and there is no direct relationship between the supplier and the user. Without tight control of the path you can’t reflect the price of a given infrastructure element back to the user. Isn’t it strange that dedicated fiber is cheaper than sharing a path? The solution is not to charge the highest price you can – the solution is to make capacity available and let the users discover what works and what doesn’t.

Today we’re confused because we keep trying to treat the Internet as a TV channel and it works far better than it is “should” so instead of buffering and using peer caching we find we can just stream video content. And then we get surprised if sometimes that doesn’t work as well because we don’t use algorithms that degrade gracefully.

The bigger problem is that we don’t see the value of the Internet as basic infrastructure and don’t see the high value in the myriad other was to take advantage of connectivity. Yet we keep worrying that we might use up the Internet.

PS: I actually do care what it costs to get bits between here and Timbuktu because it acts as a punitive tax on developing economies (and they should all continue to develop). We can’t afford to keep the worlds’ economies captive within houses of glass.

Five years ago I had a vision: the Metropolitan Internet.
And I had the same idea about how the Internet bandwidth was wasted when looking for local websites.
How wonderful if when I am looking for something local my request wouldn't travel around the world, but would use the local infrastructures, making it easier, faster and cheaper.
I am glad to see I was not wrong.
Post a Comment