Updated 12/8/2007 - The EFF (Electronic Frontier Foundation) last week publicly joined Free Press and Public Knowledge in recommending a metered Internet service as the alternative to Comcast's BitTorrent throttling. The extremist "Net Neutrality" crowd that wants to regulate the Internet with bans on per-user charges/contracts for Enhanced QoS are so busy trying to revive their cause by using the Comcast issue that they're overlooking the fact that these three groups are trying to bring you a metered Internet service. The media for the most part has missed the boat on what's really going on and they present this to the public as if EFF is trying to protect the public's interest from evil corporations.
The EFF goes as far as touting the Australian model for broadband service. Just to be sure this isn't some kind of mistake, I personally confirmed with EFF this is what they want. In their report they write:
The Australian broadband market offers an illustration of how this can work in practice. The selection of Australian broadband options can be searched at http://bc.whirlpool.net.au/bc-plan.cfm. It includes a wide selection of plans with different peak and off-peak quotas, some with a traffic shaping after a quota has been passed and others with a wide range of per-gigabyte fees. It also includes explicitly "no set limit" plans where the ISP reserves the right to deem certain usage excessive, and more expensive, truly unlimited plans where the user can saturate their link 24/7 if they wish.
I checked out the link and a Cable broadband connection costs $40/month with a 400 MB cap and a $150/GB overage charge. Just imagine if you accidentally left the BitTorrent client on for a weekend or if the kids use Grandma's computer to download a bunch of videos racking up hundreds of dollars in charges. We're all going to have to go back to the cell phone model where we worry about peak and off/peak hours and how many megabytes we used just like we worry about how many minutes we use.
Well no thanks EFF, I as an American have no interest in paying higher prices like they do in Australia (no offense to the beautiful country of Australia and its people). Not only does a metered Internet service plan screw the low-end users, it makes BitTorrent or any kind of peer-to-peer networking cost prohibitive. The EFF ironically claims its standing up for BitTorrent rights when it fact it would kill it with metered Internet services.
The article incorrectly states that EFF endorses legislation or regulation that would force ISPs or users to offer only metered services. The EFF report actually states that the *availability* of metered access alongside "all you can eat" plans, combined with accurate advertising by ISPs, is one alternative that might solve whatever congestion issues Comcast might be having (as the language you quote in your article expressly makes clear).
Nowhere in this blog post do I state EFF would force ISPs to *only* offer metered services? All I said was "The EFF goes as far as touting the Australian model for broadband service" as a better alternative to Comcast's current model and I included the Australian ISP link the EFF pointed to. The plans that came up were mostly metered plans and some were very expensive unlimited plans. Peter Eckersley even sent me an email touting this page where you pay $65/month AUD for a plan that gives you 8 GB of "pre-paid data" during noon to midnight [Update 12/12/2007 - Peter Eckersley emailed me saying he sent me the wrong link and had meant to link to this page which is $20 cheaper. That's slightly better but the 8GB cap is still a horrible idea]. Since you can download 8 GBs in less than 2 hours at 10 mbps, you essentially give up using any BitTorrent from noon to midnight unless you want to pay $3/GB. Even the off-peak rates are metered so you still have to be careful to turn off your BitTorrent client after 1 hour each day. If you want 48 GB "pre-paid data", you need to pay $120/month AUD and $3/GB over that amount.
Now consider Comcast's offerings which permit you to download and upload unlimited data using BitTorrent with no throttling for a flat fee of $40 per month. You can easily download 100 GBs and upload 10 GBs per month or more and Comcast won't stop you or charge you anything extra. The only thing Comcast does is occasionally scale back the number of BitTorrent seed connections (dedicated server mode) you can have even though Comcast's TOS (Terms Of Service) prohibits servers of any kind. My ATT DSL plan is less than $20/month and I can download 8 GB per day every day and not pay a single cent on overage charges so what is the EFF thinking recommending the Australian ISP model over Comcast's "bad" model?
The EFF says what Comcast is doing is evil and that the Australian model is the better alternative even though it's draconian compared to what Comcast or any other American ISP is doing. It would certainly stop the BitTorrent usage during peak hours but at what price to the user? The Free Press and Public Knowledge also think metered Internet is a better alternative but they go a step further and want to criminalize Comcast's current operating model and fine them trillions of dollars. So again I ask: Who is the EFF, Free Press, and Public Knowledge serving? The RIAA and MPAA couldn't buy this kind of anti peer-to-peer lobbying if they tried.
<Next page - My phone debate with the EFF>
The EFF which came out last week to publicly support the Free Press and Public Knowledge read my report "A rational debate on Comcast traffic management" and they weren't pleased. So when Peter Eckersley of the EFF and co-author of their last week's report on the "Comcast Affair" sought me out a few weeks ago via email and requested a phone conversation, I called him back the same day. Eckersley tried to convince me that Richard Bennett - the computer networking pioneer and practicing engineer who explained the technical details in my Comcast blog - was all wrong about Comcast's network congestion problems and that he, a "computer scientist", knew better. When I asked Eckersley what his background was, he explained it was in "copyright law".
Before we got in to the conversation, I wanted Peter Eckersley to clarify the EFF position first so I asked if the EFF supported the Free Press and Public Knowledge position that demands an immediate FCC enjoinment BEFORE the facts are examined. Eckersley basically explained that he and the EFF are so confident in their alternative network management mechanism yet-to-be-tested theory that immediate action without examination of the facts was indeed warranted. Since Eckersley calls himself a "computer scientist", I asked him what the scientific method was and Eckersley explained that science was really philosophy (presumably his). Then in a bit of doublespeak that would make George Orwell envious, Eckersley explained to me that "no action was an action". Therefore, we couldn't take the "no action" action before the facts are in and the Government must force Comcast must adopt his theories immediately.
After a bit more heated debate, Eckersley eventually figured out that I wasn't going to sing to their tune so he told me that I should just "shut up" if I wasn't going to help their cause. But when lawyers and bureaucrats who sue people for a living try to make public policy decisions for the rest of us and sell us down the drain, the last thing I'm going to do is "shut up" and I'm going to make sure that the public hears about it.
<Next page - EFF's alternative method for network management>
There are methods available to Comcast to limit the amount of traffic that P2P software transmits on their network, without preventing any categories of connections, interfering with any protocols, or forging packets. For example, ISPs can implement dynamic per-user traffic shaping. They can set a limit on the amount of data per second that any user can transmit on the network. They can also set these limits on a dynamic basis, so that (1) the limits are gradually relaxed as the network becomes less congested and vice-versa and (2) so that the limits primarily slow the traffic of users who are downloading large to very large files that take minutes to transfer. We have observed Comcast to take most of these steps in managing their cable networks, but in our testing, we have never seen them make the kinds of dynamic adjustments to their rate limits that would be necessary to gracefully avert severe network congestion (23). This suggests — though it cannot prove — that even if Comcast began forging RST packets in response to problems with network congestion, they did not exhaust the reasonable, user-friendly, and standards-compliant responses before they began taking decidedly less reasonable measures.
So the EFF based on their "expertise" with zero test data or experience in network management wants to tell Comcast how to run their network. Eckersley and the EFF tells us [UPDATE 12/8/2007
that the James J. Martin and James M. Westall paper referenced by Richard Bennett that Richard Bennett's interpretation of "The Interaction Between the DOCSIS 1.1/2.0 MAC Protocol and TCP Application Performance" explaining the adverse effects of BitTorrent on DOCSIS networks] is wrong even though they lack the credentials of Martin, Westall, and Bennett. The efficacy of EFF's solution is questionable since throttling thousands of connections per second with traditional traffic shaping mechanisms may have little effect on the first-hop DOCSIS network topology. Even if the EFF, Free Press, and Public Knowledge alternative did work, the cost of implementing such a system would be massive compared to what Comcast currently implements.
Since there simply is no traffic shaping mechanisms in DOCSIS 1.1 cable modems which are currently deployed to millions of users, the traffic shaping load would be entirely foisted upon the router infrastructure. Since we're talking about a dynamic throttling policy that has to take in to account application state and individual BitTorrent sessions of millions of users each of which open hundreds of sessions per hour, the existing routing infrastructure would require a massive infrastructure upgrade. Anyone with experience in routers and firewalls will know that the bigger an ACL (Access Control List) is the slower the router and firewall becomes since each packet that gets forwarded in a router has to scan through the entire ACL. Most routers handle ACLs that may be tens or hundreds of lines long, so imagine the kind of router you need if you wanted to have an ACL a million lines long that has to be dynamically updated with hundreds of thousands of new lines per hour.
Comcast's method of handling traffic management involves using a device (believed to be Sandvine) that sits on the network which shoots down connections on its own. So instead of injecting thousands of dynamic ACLs in to the routers per second bringing the routers to their knees, the easiest and cheapest way to kill off excessive peer-to-peer connections to prevent a network meltdown is to ask the client to stop transmitting instead of asking the routing infrastructure to block it. Since BitTorrent has no such congestion control mechanism and it has explicit design goals of bypassing detection and ISP throttling not to mention copyright restrictions, the only machine language it understands is a TCP RST (Reset).
The EFF calls this packet "forgery" but this type of technique is common in the networking and software industry where alternatives don't exist. NAT (Network Address Translation) for example is commonly used by ISPs to handle the IP shortage and they forge the IP addresses so that multiple users can share the same IP. The EFF wants you to believe this is like Comcast forging a letter from your loved one saying they don't want to talk to you anymore but in reality, this is just a congestion control mechanism that prevents the network from melting down. It doesn't prevent you from using BitTorrent; it only prevents you serving too many users as a BitTorrent seed during peak minutes which already falls under Comcast's "no server" policy. I spoke with Comcast and they explained that these peak times may last a few minutes and they use have to use the TCP resets to stabilize the network. This isn't content- or protocol-based "discrimination", it's complete content and protocol neutral and clearly falls under the category of reasonable network management.
It doesn't need to remain like this forever since Comcast is already talking about upgrading to DOCSIS 3.0 in 20% of its network by the end of 2008. Richard Bennett explained to me that the newer cable broadband protocol should obviate the need to use the TCP Reset "kludge" since DOCSIS 3.0 cable modems have traffic shaping built in to them. So the market pressures from Verizon's superior fiber optic FiOS network will simply force Comcast to improve their network infrastructure and none of these draconian measures from the EFF are necessary and trying to force a DOCSIS 1.1 cable modem to behave like a DOCSIS 3.0 modem simply doesn't work. If the EFF, Free Press, and Public Knowledge gets their way, the costs sky rocket and you the consumer gets stuck with a higher bill via metered Internet service.