Netflix Inc. May Have Found the Magic Bullet to Kill Comcast's Connection Fees

Encryption doesn't matter, all they need to do is traffic shape everything that's not "web" (by that I mean strictly 80 or 443) and they can effectively kill any peer-to-peer thing that netflix could possibly come up with, while still maintaining an illusion of good service for 99.9% of the rest of their customers' traffic. It would deprioritize bittorrent too, so two birds with one stone as far as Comcast goes.

Let's face it: almost everything an average internet user does, measured by number of packets, is on 80 or 443 (except bittorrent, but comcast very likely already deprioritizes that.) Maybe a sprinkling of email traffic here and there for the few who don't use a web mail provider. I'd actually be quite shocked if Comcast didn't have traffic shaping in place that deprioritized non-web traffic. It would make smart business sense.

Which means the only way Netflix could do P2P is to put peers behind port 443, which can only work for one connection per peer. (And good luck teaching all netflix subscribers how to get NAT working on their routers.) Or they'd have to do a nat busting service, which wouldn't be peer to peer any more, and would open them right back up to Comcast's throttling.

At the end of the day, if nobody's forcing Comcast to stay neutral, they can deprioritize whatever traffic they want. Who's going to stop them? What are you going to do, sign up for a competing ISP?

And what would prevent them from running port 80 and 443 cross-domain p2p ? If I remember correctly, kazaa had a mode that worked like this.

Suppose you'd do it really, really simple. If you watch a netflix show, you download an encrypted version, and start serving it on your IP address on ports 80 and 443, http and https (wouldn't work with NAT of course, but ...). Then inform central servers that this link exists and the central server tells 1/10000 of next requests to download it from you. Now split up the files into 1000 pieces each, and make it work like that. It could take care of tracerouting to everyone and try to "group" requests with the most overlapping traceroutes.

Bittorrent is a brutally simple protocol, compared to it's predecessors, and I'd say, even compared to things like http and https. Running bittorrent over http or https would not be very hard.

Keep in mind that there's also an advantage for the providers. What the providers really want is payment for building network infrastructure. Given how hard that is becoming, and the fact that it's not linear (for linearly increasing endpoints and bandwidth you need exponentially increasing network equipment and transfer lines), they should be ecstatic about efforts to turn traffic back as soon in their network as possible, it would offload their core networks, and they can sure use that.

Contrary to what people think here, in most cases ISPs only throttle p2p traffic either between their distribution and core layers, or between their core and peering (second makes more sense financially, but it may not be the same organization). So p2p traffic between you and your neighbor is not throttled at all. This is done because firewalls that can reliably separate p2p traffic from other traffic are expensive, so one tries to avoid sending all core traffic through them. Throttling p2p in the access servers is quite expensive, and will give customers (even more) the impression that their line is throttled at all times.

(also in most ISPs even a basic business connection will bypass p2p throttles, given that the cost difference between customer connections and basic business connections is little or nonexistent these days, you may want to consider switching)

This article seems quite naive. Painting P2P traffic as something "which Comcast cannot block" ignores both history and the reason why Netflix paid for direct peering with Comcast in the first place. It's not that their traffic was being "blocked", just that it was coming in on the same overloaded pipes as other general Internet traffic. While Comcast customers might not notice a web page taking a few milliseconds longer to load, they certainly notice the difference in video streaming quality, and hence why the new peering arrangement made a difference.

Shifting all the traffic to P2P most likely would send Netflix traffic back into the overloaded slow lanes again - and if anything makes it all the MORE likely to be further slowed down and deprioritised for reasons. Phrases like "Internet traffic of a type often used for piracy" are all that needs to be said to justify whatever further shaping, blocking, and other indignities could be applied to it.