First of all, thank you for joining me in my call for more ISP transparency.

Second, thank you for dropping the idea that “up to” somehow equals “at least”. There’s been too much of an expectation that the “up to” bandwidth somehow means a minimum guarantee. I’m glad that you’re finally able to come around and accept the fact that Broadband will always have contention ratios.

As far as I’m concerned, I’d be very happy to have 1000:1 contention ratios if it means that I am sharing a 1000 Mbps pipe. Statistical multiplexing and some sort of fair share algorithm would mean that I get to burst to 1000 Mbps and get my short requests more instantaneously. Larger pools also mean better statistical stability. Statistical multiplexing and fractional ownership via usage caps allow the flexibility to boost peak performance and that ultimately benefits the ISP customer.

As I explained in this blog posting, there are two things you should really consider fixing in your proposal.

1. You need to be careful that you’re not advocating that contention ratios must be no more than 2 to 1. The way you’ve set up your example seems like you’re advocating no more than 2 to 1 contention. My suggestion to you is that you clarify it such that it means 50% (or whatever the ISP wants to set and advertise) minimum performance as an hourly average. That allows for temporary drops to 10% performance but they can make up the averages over the course of an hour.

2. You should specify maximum permissible latency that is not self induced. The way you’ve worded your proposal means that anyone who runs a P2P application will instantly jack up their own latency and jitter and cause the ISP to fail in their service delivery. That’s not a reasonable requirement for the ISP.

Furthermore, this is all the more reason that ISPs should be priorizing low bandwidth applications (low could also mean low average bandwidth e.g., web browsing) over high bandwidth applications. That allows P2P to be less toxic which means fewer people shutting off their own P2P client which translates into more seeders. Since I know you’re an avid P2P user, I invite you to join me in calling for more prioritization http://www.digitalsociety.org/2009/09/fcc-5th-principle-must-allow-for-reasonable-discrimination/.

The ISPs currently are able to come up with a number and emblazen it on all of their ads. Some of us actually get that number, but many do not.

If we were to take the true contention ratio, we’d be diving the size of the bandwidth pool by the number of homes passed who subscribe to Internet and we’d have both the accurate speed and an amazingly small speed. That’s fair, but it truly isn’t what the ISPs are selling because we know that all those subscribers won’t be on at once and that things like bandwidth consumptions limits and other traffic management techniques can further improve the minimum speed that an ISP can virtually guarantee.

The guarantees of speed, latency, and uptime are to the ISP’s border routers or access routers (where their ‘net meets the rest of the ‘net). We think it’s important that the ISP be able to set its own service goals and raise the bar on itself and its competition. While it is useful to empower users with stated minimum expectations, we’re flexible as to the fine details (how long is ‘extended’?, how is this reliably measured?, and etc.) because ISPs need to manage some customers who have unrealistic expectations from the starting gate. The idea is that few customers will complain if ISPs generally treat them right in an area that is grossly oversubscribed, but there’s room here to bicker in the face of the few remaining hard-corps serial complainers.

There’s still room for improvement on the idea. Now that you know the goals, I invite you to help.

Robb Topolski
Chief Technologist
Open Technology Initiative, New America Foundation