Well that certainly didn’t take long.
Richard Bennett has an article at The Register describing BitTorrent, Inc.’s new method for circumventing traffic throttling. Essentially (if I understand it), BitTorrent has altered the way in which its uTorrent P2P application will work. Instead of relying on the Transfer Control Protocol (TCP) uTorrent will now use the User Datagram Protocol (UDP) to move packets. Richard describes what this means and the potential impact of this better than I can. Critically, however, Richard describes this as a means by which BitTorrent can avoid Bell Canada’s targeted traffic management by disguising the nature of its traffic as latency-intolerant (like voice over IP (VOIP))and therefore given priority over other traffic. You can see some discussion of this as a response to the CRTC decision to allow Bell Canada to manage traffic here at DSL Reports.
As I observed only last week, the CRTC decision presents a splendid opportunity to grab some popcorn and watch some other country play games with its critical infrastructure. Mind, since the internet is a global “network of networks,” what happens in Canada is likely to impact me here in the U.S. as well. But I can’t do anything about that. So pardon me whilst I munch my popcorn and enjoy a good dose of Cassandrafruede (a term of my own invention which means “the bitter pleasure experienced when something awful you predicted that could have been avoided if people had listened to you comes to pass, even though you also get screwed through no fault of your own”).
More analysis to go with my popcorn below . . . .
This sort of “arms race” between application providers and network operators was one of the predicted outcomes by pro-net neutrality advocates during the Comcast/BitTorrent fight. Indeed, David Clark — who absolutely hated the concept of government getting involved in regulating how network operators managed traffic, fretted over this at the FCC Boston hearing on Comcast/BitTorrent in February. Clark feared that Comcast’s use of reset packets to interrupt file transfers as a means of congestion control would prompt a widespread distrust of the underlying protocols that govern internet traffic and would prompt application providers and network operators to engage in a non-stop escalating war until we achieved fragmentation of the internet.
So I’m not surprised by the efforts of BitTorrent to route around Bell Canada’s traffic management scheme. It is the natural market response in a deregulated environment, and those who applauded the CRTC for “getting right” have no call to complain about BitTorrent’s conduct anymore than they had about Bell Canada’s. Richard Bennett correctly points out in his Register article that “the internet is only a stable system because application developers are gentlemanly with regard to the amount of traffic they shove onto the network.” But that misses that — like any good relationship — the obligations run both ways. Application providers can be “gentlemanly” when they think the underlying system works in a reasonable and predictable manner that gives them an equal shot with anyone else. If network operators have no responsibility to play by the old rules or to work with application providers, application providers will – predictably! – take matters into their own hand. Any other outcome is tantamount to suicide for the application provider.
Similarly, the user encountering network policies or network management techniques that restrict traffic in unknown ways feels not the least bit of obligation to observe these or responsibility for the maintenance of the system as a whole. Does an SUV purchaser feel any concern about the “fair” distribution of gasoline or the possible impact on everybody’s climate? And why should a rational user, who finds himself or herself subject to arbitrary regulations the provider may change at will and that the provider may not even disclose, respect these limits? Users buy a service (access) because they want something. If the access provider screws up the ability of subscribers to get what they want (here, BitTorrent downloads), it is a predictable certainty subscribers will act to the best of their ability to use the service they are buying in the manner they want. Crying about their ignorant and selfish behavior, or lambasting application providers as serpents in the Walled Garden of Eden the network operators want to provide, is as effective a response as lecturing snowmobile users about the need to preserve our national parks for the next generation so please stop blasting around all over the place on machines you bought for the express purpose of blasting around all over the place.
The next predictable response will be for Bell Canada to take further steps to clamp down on traffic management. This may take several forms, from efforts to do source-based blocking/degradation as a means to block p2p applications using UDP, or it may trigger bandwidth caps as the alternative means of congestion management. I confess this later will give me huge Cassandrafeude jollies, as one of the primary objections to the FCC finding for us was that it would trigger bandwidth caps (although the evidence that the FCC’s Comcast/BitTorrent decision prompted bandwidth caps seems rather dubious in light of the fact that Comcast had already implemented them (without disclosure) before we filed the FCC complaint).
Or Bell Canada may come up with some different response. Network operators are quite inventive. Given freedom to manage their networks without supervision, I am confident network operators will come up with some kind of solution that solves their perceived problem. Sadly, I also predict the solution the network operators come up with will shift costs to others, who will then respond with their own “fix.” All of this will have continued, steady impact on the underlying ecology of the internet. Again, I will experience some huge Cassandrafruede jollies if the ultimate outcome of the private management of the internet is the fabled tragedy of the commons rather than the Promised Land of Ronald Coase.
For me, therefore, the BitTorrent response to Bell Canada provides some rather quick confirmatory proof that (a) “the free market” really does work in predictable ways, in that those with incentive and ability to act will do everything they can to maximize their short-term advantage heedless of the possible damage to the underlying system in the long run; and, (b) most techno-Libertarians seem unwilling to believe that the Gods of the Marketplace will permit such blasphemous outcomes to occur — and will therefore see the bad outcomes as the result of the evil application providers abusing the innocent network operators rather than as a predictable market response. But as we members of the Congregation of the Progressive Capitalist understand, “The Market” does not care about outcomes. The Market moves in accordance with its own dictates, indifferent to the results, bringing highly predictable ruin even unto itself if not properly managed and directed as needed. Which is why I “trust the market” but I do not worship it.
You want certain outcomes? Be prepared to regulate where reasonable understanding of market incentives predicts that the most likely outcome is not what you want. Otherwise, take what comes from the Gods of the Marketplace, who remain both predictable and utterly indifferent to your needs or desires.
Stay tuned . . . .
Not surprised. What I am surprised about is that the bittorrent folks haven’t really bit the bullet and applied some sort of ssh or vpn scheme on top of the torrent protocol to really freak the ISP’s out. masking their signature under something that everyone uses for legitimate traffic forces the problem into the ‘too tough’ bucket for the engineers.
Harold, one other observation. I might question Bennett’s article somewhat. SIP utilizes TCPIP for build up and teardown of a call and can use either TCP or UDP to carry the audio. My experience is that its 50/50 based on provider. Skinny uses TCP for build up tear down and UDP to carry audio. But skinny is generally associated with Cisco CallManager type installations. So though the call maybe UDP inside the facility generally once outside the facility its transported over TCP or TCP-VPNd.
VoIP maybe affected less by this than the likes of meeting services providers like WebEx, GoToMeeting and DimDim. Their services are highly leveraged on UDP.
But you know where ever there is a problem there is a profit to be made. The bittorrent folks ought to get together with the carriers and work out a scheme. They would develop code that makes it easy to skim the traffic. The carriers open a torrent channel in their networks. The customer pays extra for the privilege. Really no different in one sense than paying for a premium channel. The carrier and the bittorrent split the profits accordingly.
The quote from Bennett about gentlemanly developers misses another large user of the internet: email spammers. They operate using a TCP-based protocol (SMTP), and on some backbones can eat up 20% of bandwidth. I only note that providers have not had much luck sorting good SMTP traffic from bad at the backbone point, and wonder if they will be successful throttling BT (and other p2p) in general.
An interesting note about the article in the register. UDP is *not* designed for real time protocols. It’s designed for applications where packets can appear more randomly, and that the application will control error checking & retransmitting. That starts to get towards the very core of P2P – the app deals with errors, and breaking things apart.
Actually, UDP was designed for real-time applications for which the TCP retransmission strategy is not appropriate. Other uses are legal, but that was the goal.
@Matthew: On the UDP point, go here: http://www.reed.com/blog-dp…
Then admit your error and pray for mercy.
Harold, I actually do agree with you on one point: BitTorrent will, indeed, be “disguising the nature of its traffic as latency-intolerant.” Or, to put it more accurately, it will be MISREPRESENTING it as such. (Bulk file transfers are not intolerant of latency OR delays.) Those who claimed that the use of RST packets to manage traffic is “forgery” (even though it has a long history of legitimate use) will have to admit that this misrepresentation is far more blatant.
You might enjoy my follow-up:
A lot of questions remain about the soundness of the BitTorrent approach.