Tuesday, May 02, 2006

the myth of the neutral net

While I am sympathetic to the notion that The Net Should Be Neutral, I have a hard time seeing how the recent changes present existential threats to the Net As We Know It. The Net has never been neutral. It is hard to imagine how it could be.

Let's start from the most obvious point, the connection from a server to an ISP. Half the country still connects to the internet via dial-up lines. Since these connections are not always-on, systems behind those lines are not fully part of the Net.

People get around that by subscribing to hosting services. That's great, but hosting services price by the level of service. Anyone who's tried to run a website then had it get too popular (instapounded, atriated, fark'd, slashdot'd, etc.) knows that you have to pay more to reach a wider audience.

More subtly, the Net is not an amorphous blob of bandwidth. It's a series of point-to-point connections. The route from my system to cnn.com has at least 16 hops. It has 80 ms of latency. Yahoo.com has 10 hops and 20ms latency. If I do the same test from our family website, I see 13 hops and 75ms latency to cnn.com, and 11 hops and 80ms latency to yahoo.com. I suspect high-level ISPs already compete with one another to provide the lowest latency and highest bandwidth to the broadest area. The only way to eliminate such biases would be to eliminate competition between ISPs.

In truth, it gets more complicated than that. Some sites are geographically mirrored. Some aren't. Some are cached. Some push content out through dedicated content distribution networks so that the bulk of data is close to the client. All of these approaches require more expertise and money than casual internet users are able or willing to muster, and there's almost nothing that can be done to prevent them.

So if companies start differentiating between their customers, it's not as if they're destroying a level playing field. The field has never been level and won't be until bandwidth, storage, and computing are free.

Such practices may make things marginally worse, but they may also make them marginally better. Improving service today is expensive. If backbones commoditize service priorities, they should be less expensive than the approaches described above. If they're less expensive, they'll be available to a wider range of customers. Upgrading your web hosting service could include upgraded backbone service as part of the package.

I'm not a pollyanna about this (or much else, I guess), but on my potential end-of-the-world-as-we-know-it scale, I'd have to give this no more than a 1% Abu Ghraib.