The Defense

Whatever the case, Comcast has robustly defended its treatment of P2P traffic. And so have others. In the pages of The Register, network architect Richard Bennett has argued that BitTorrent puts an unreasonable strain on Comcast's network - and that the company's best option is to clip the app's wings.

Bennett points out that BitTorrent runs "dozens (or even hundreds) of TCP streams concurrently" - and that other apps typically open only one. "Fundamentally, this Comcast-BitTorrent issue isn't real complicated," he says. "What it comes down to is the fact that the BitTorrent user is essentially transmitting several times more data upstream on the Comcast network than anyone else."

He acknowledges that at any given time, BitTorrent uses only a small fraction of the TCP streams it opens. And he agrees that Comcast could put a general cap on each user's upload traffic - without specifically targeting BitTorrent. But he's adamant this method doesn't suit a cable network. "Eventually, it would have the effect of reducing the amount of traffic that BitTorrent offered on the congested upstream link, but it would also be slower and much less effective."

And in the end, he believes it doesn't matter how Comcast deals with BitTorrent. "If the fundamental problem is that you have a class of user who are bandwidth hogs, and if we recognize the ISP has a legitimate responsibility to prevent bandwidth hogs from consuming so much bandwidth on their ISP network that other people can't have a good internet experience, do we really care what method they use to fulfill that responsibility?"

Bob Briscoe, chief researcher at the BT Network Research Centre, has no interest in taking sides in the Comcast-BitTorrent debate. But he will say that capping overall upload traffic on a cable network is more problematic:

The problem is there's no equipment that's under the complete physical control of the cable operator and located before upstream traffic starts to share the cable network used by other customers.

So instead of physically limiting traffic on entry, most traffic management solutions limit traffic half-way across the network. When such boxes drop packets from p2p data flows, the transmission control protocol (TCP) software in the operating system of the sending user's PC assumes there's congestion and slows down voluntarily.

If the sender's OS didn't voluntarily back off, the heavy user would still swamp other users on the cable network. All the offending traffic would be dropped once it got to the policer - but too late to have prevented any harm to others.

Fortunately (currently) nearly* all p2p solutions use choose to use TCP to control their transmission, which responds voluntarily to perceived congestion.

*[I know of a couple that don't]

Yes, he says, an operator could position some sort of policing equipment closer to the end user - but this would be expensive. And if a cable operator does find a way to effectively cap each user's upload traffic, Briscoe adds, it faces another quandary. "If they just cap all a heavy users' traffic indiscriminately, the non-heavy apps (Web etc) of a heavy user also slow to a crawl."

Of course, Comcast has now said that a fix is on the way.