Even if the FCC's Comcast order had been based on sound legal and technical footing, its implementation would require extra-human powers of perception since the Commission couldn't be bothered to lay out precisely which features of the Comcast system it finds objectionable. The order criticizes the use of TCP Reset packets to prune the thousands of connections a P2P application can create in a day’s time, but doesn’t ban the technique. That's fortunate, because home gateways also use Resets to clean up NAT tables.
The order objects to the application of bandwidth quotas to classes of applications, but its signatories have issued press statements endorsing priority boosting to benefit other classes of application, such as VoIP.
The only management techniques the order says an ISP can apply that clearly won't run afoul of future whole-cloth enforcement actions are the most heavy-handed. An ISP may apply volume-based billing, the worst possible method for P2P companies, or it can "throttle back the connection speeds of high-capacity users," the worst possible scenario for home network users.
The flat-rate pricing model we’re all used to only works when users consume roughly equal amounts of bandwidth, but the adoption of P2P for non-stop use by a small number of people blows it out the window.
Comcast has adopted a bandwidth cap of 250GB to prevent just one per cent of its customers from consuming half its download capacity. Japan's NTT has adopted 900 GB monthly upload caps on its fiber, Time-Warner is experimenting with metered pricing, and Frontier Communications has adopted a full-scale metering regime that prices downloads at $4 per gigabyte.
When ISPs are forbidden from de-prioritizing P2P streams, management-by-pricing is inevitable.
The only sensible method of traffic management that's ever been proposed for home networks is a two-stage system that does the following:
Firstly, it divides it bandwidth equitably among users at a given service tier; and then, secondly, it allocates delays among the streams to and from each user account according to application needs.
This approach is necessary because internet users do more than one thing at a time. When someone in the house is using VoIP, someone else is surfing the web and both may be downloading files using P2P. Under the FCC's new system, this sort of mixed-use scenario qualifies the entire account for connection speed downgrading, which can only be overcome by purchasing a higher service tier.
Save the Internet is already complaining about the price increases their petitions have wrought.
"It is a false choice to suggest that since Internet service providers cannot arbitrarily block online content, they will be forced to meter. There are a whole host of other non-discriminatory options available to providers that are more effective at managing congestion," the group says in a statement.
Unfortunately, there are no management approaches for mixed-use scenarios that don’t involve just enough discrimination to identify application needs, and the FCC has just banned that practice.
Thanks a lot, folks.®
Richard Bennett is a network architect and occasional activist in Silicon Valley. He wrote the first standard for Ethernet over twisted-pair wiring and contributed to the standards for WiFi and the Ultra-Wideband wireless networks. His eleven-year old blog is at bennett.com.
@ Andy Worth
"RE: Because of the Freetards who leech off people like me (who pay for music)" -Sorry, but if you don't want them to leech off you, stop sharing your music collection online. "
Sorry you don't get it. If ten units of a product are manufactured, and nine are not paid for, then the person who buys the tenth is subsidizing the other nine. If you run a shop, and half the stock walks out of the door without being paid for, the shopkeeper must charge the honest, paying customers more than they would otherwise pay.
So when freetards leech, the cost of music goes up for everyone.
I'm sure you're a good lad who never downloads anything illegal, but your economics isn't so hot.
Oh Canada (hello Johnny Boy and Mr Mezei)
Actually JohnnyBoy, I read Mr Mezei's submission some while ago :)
Unfortunately he and I have different definitions of "deep packet inspection", and afaict there is no industry-agreed definition (probably deliberately), so please let me explain my position.
If my ISP wants to know whether my packets are http from the BBC or iPlayer from the BBC or ftp from Novell Inc or streaming from Youtube or wherever, I have no problem with that. Maybe JF does, that is his privilege, and it's fine by me.
My ISP uses their knowledge of their customers' traffic types and sources/destinations (NOT PACKET CONTENT) to keep customer costs down whilst providing affordable service acceptable to many. An ISP without "traffic management" would probably charge me at least 50% more per month, in order to maintain adequate (unmanaged) free headroom on their network to cope when peaks in demand occur.
Some folk may consider that extra 50% a worthwhile spend, in order to avoid "traffic management". But that extra 50% won't get them "unlimited" use, just a network where traffic must be managed manually (and whose users must be managed manually) to avoid the network getting congested as demand increases. If that's their choice, it's fine by me.
However, if my ISP wants to know the content of the web pages I have been reading, that requires analysis of packet *content* as well as IP addresses and protocol types and such. No sensible Internet user likes that kind of "deep packet inspection", although equally no sensible Internet user should assume they have any guarantee of privacy on the Internet.
In the UK, most people have a choice of retail ISPs (even though BT, the former monopoly telco, is dominant, and their wholesale arm, BTw, is the only wholesaler offering national coverage, and is thus the only option available to many smaller quality and service focused ISPs from AAISP to Zen). Some retail ISPs offer a quality (but expensive) service without "traffic management". Some big-name cowboys have traffic management (and at least one has had Phorm too). Choice is good (Phorm isn't).
I do as little business with BT as I can, even before Phorm. I use an ISP whose use of "traffic management" aligns nicely with my needs and whose tariff structure nicely matches my needs and budget. Those who want an ISP with no "traffic management" either have to pay more (maybe a lot more than I'm willing to pay) or be lucky enough to be in the half of the UK where there are alternatives to BTwholesale's ridiculously-overpriced services.
On a related note, JF's submission refers to what happened when the dial-up ISPs ran out of capacity - in those days, the limiting factor was not bandwidth, but the number of dial-in ports available. In Canada, he says they installed more capacity ie more ports. I have no idea whether it's true, but I know that here in the UK, the port capacity "problem" wasn't solved by adding more capacity (which would have cost money), it was mostly "solved" by limiting dialup sessions to a maximum of two hours (or something like that). In other words, the ISPs didn't install more capacity, instead they took measures to manage demand. It wasn't the end of the world, even though the flat-earthers at the Campaign for Unmetered Telecom had said it would be. And metered broadband, and traffic management, won't be the end of the world either. Wide deployment of Phorm-type technologies will be, and on that hopefully we can agree.
Define the service first
Good article, there is a real danger in defining fair but bad non-discrimatory network management practices. We do not need the equivalent of PSTN call blocking. Tell me what I am paying for? what resources have been built to support the service I pay for!
I think the FCC should focus on apply existing trade description legislation so there is more service transparency. I have had a go at a label here. http://bbbritain.co.uk/kitemark.aspx
Once there is more transparency of service parameters, consensus on best congestion management practices could then be progressed.