A large group of exceptionally clueless souls got together recently and decided to bare their ignorance to the world, just in case anyone was in doubt. They engaged in the one of the great, time-honored practices of puffed-up egos everywhere and wrote an ersatz piece of “model legislation” dictating that anyone who doesn’t understand “The Internet” as they do is bad and wrong and can’t call his product “The Internet”.
Gosh, what would lawmakers do without such expert advice? Here’s their deep insight:
SEC. 3. DECEPTIVE PRACTICES IN PROVIDING INTERNET ACCESS.
(1) Definitions.- As used in this Section:
(A) Internet.- The term “Internet” means the worldwide, publicly accessible system of interconnected computer networks that transmit data by packet switching using the standard Internet Protocol (IP), some characteristics of which include:
i) Transmissions between users who hold globally reachable addresses, and which transmissions are broken down into smaller segments referred to as “packets” comprised of a small portion of information useful to the users at each transmission’s endpoints, and a small set of prefixed data describing the source and destination of each transmission and how the packet is to be treated;
ii) routers that transmit these packets to various other routers on a best efforts basis, changing routers freely as a means of managing network flow; and
iii) said routers transmit packets independently of each other and independently of the particular application in use, in accordance with globally defined protocol requirements and recommendations.
(B) Internet access.- The term “Internet access” means a service that enables users to transmit and receive transmissions of data using the Internet protocol in a manner that is agnostic to the nature, source or destination of the transmission of any packet. Such IP transmissions may include information, text, sounds, images and other content such as messaging and electronic mail.
(2) Any person engaged in interstate commerce that charges a fee for the provision of Internet access must in fact provide access to the Internet in accord with the above definition, regardless whether additional proprietary content, information or other services are also provided as part of a package of services offered to consumers. [emphasis added for laughs]
That’s heavy, isn’t it? There are only about five blatant falsehoods and two obvious contradictions in it.
My favorites:
1. They act like the Internet is a network, and not simply a means of interconnecting networks. The Internet doesn’t actually care what you do on your private network, it only sees the things you pass to other networks.
2. They ignore the structure of the IP header, which includes a TOS field with several levels of priority.
3. They don’t understand the fact that routing depends on commercial contracts, the enforcement of which depends on examination of addresses and labels.
According to this exercise in intellectual masturbation there is no Internet today.
Some of the geniuses involved are:
Susan Crawford, Associate Professor of Law, Cardozo Law School (thinks the Internet is a telegraph)
Bob Frankston, Telecommunications Analyst and Visionary (actually, he’s the Visicalc guy)
David S. Isenberg, Ph.D., Founder & CEO, isen.com, LLC (fired from the phone company and mad about it)
Kevin Marks, mediAgora (has a very thick English accent but no knowledge of network architecture)
Andy Oram, Editor, O’Reilly Media (O’Reilly)
David P. Reed, contributor to original Internet Protocol design (was in the same room with David Clark 30 years ago, on a downhill slide since)
Clay Shirky, Interactive Telecommunications Program, New York University (Journalist)
Siva Vaidhyanathan, Department of Culture and Communication, New York University (wannabe Lessig)
Esme Vos, Founder, Muniwireless (lawyer)
David Weinberger, Fellow, Harvard Berkman Center (former philosophy professor)
Steve Wozniak, Co-Founder of Apple Computer, Inc., Member, National Academy of Engineers (and recently a high school Spanish teacher)
There are some others I’ve never heard of, with titles like “Librarian” and “TV Documentary Producer.”
So you’ve been warned.
It looks like they forgot to invite anyone with any experience in engineering or operating a large network to participate.
If I’m reading this correctly, they’re implying that BGP policy mechanisms are not part of “Teh Intarwebs”.
Richard, it’s pretty clear who the jerk is here, but just so your readers don’t get taken in by your ignorance, it might be germane to mention Susan Crawford’s position on ICANN, and that I worked on the most widely-deployed implementation of the IETF video streaming protocols.
The TOS header was largely unused, and subsequently redefined by the DS spec. It is of course only used within private networks, as paying attention to it’s value on entry will compromise any internal routing priority. Thus your points 1 and 2 are self contradictory.
“Circle-jerk” is an Americanism that would translate into “group wank” in your dialect, Kevin; I’m calling these people tossers, not hooligans.
Susan Crawford has a blog, and anyone with knowledge of network protocols who reads it can quickly determine that her knowledge of Internet standards and practice is virtually non-existent; her description of the Internet is indistinguishable from that of the telegraph network.
The most widely-deployed product using IETF streaming media protocols is Windows Media Player, and I wasn’t aware you worked on it, Kevin. My understanding is that you once were a member of the team that produced an early version of QuickTime, better known for its proprietary features than any standard ones. It’s my personal opinion that any video player that doesn’t support Multicast isn’t worth much. I’m happy for application programmers to opine in their fields of expertise, I’m just not blown away by their wild theories about the pipes.
The TOS header is a matter of architecture, and I’d argue with you over its implementation but what would be the point? Voice over Wireless implementations use it today, and I’ve seen this with Airopeek. Networks that use that field for DSCP typically convey the original sense in MPLS labels. Priority is very useful on packet networks, especially those with deep queues and high latency. Are you aware of DiffServ?
The points that had me laughing at wanky statement about the Internet are quite valid: packets don’t “freely route” anywhere they want, they take routes consistent with the contracts their carriers have with each other. It is not the case that any packet can take any route, the tables the policies that govern them are neither arbitrary nor magical, they’re governed by cold, hard cash. “Best effort” is simply one class of service on the Internet, not the only one. We’ve been able to buy low-jitter service from Verio, WebEx, and Global Crossing for several years, and the Internet hasn’t stopped working yet.
As you can see from the comments here, Kevin, sycophantic nonsense about Internet architecture and network operation don’t fly here because the audience is informed.
Now that you’ve had time to think about it, perhaps you can answer the question George Ou put to you about priorities Friday: how is it that QoS restricted to paying customers (which you don’t like) does more harm to your Best Effort streams than the free QoS which you (grudgingly) do like. The world is dying for some direction from you.
You mean RTSP and RTP? I can’t find your name in reference to either one of those standards.
Diffserv is not directly related to routing. It tells an interface how to handle packets when they get there, it doesn’t necessarily tell the packets how to get to that particular interface. They don’t necessarily contradict each other, but it might be confusing as routing policy does have significant impact regarding traffic flows, and you can perform some routing optimization ala Internap to greatly improve percieved Internet performance.
Anyway, this will be moot once QoS Policy Propagation via BGP becomes standardized. Once that happens, carriers can set policy on How QoS is handled on their network in a scalable fashion.
I did indeed mean QuickTime and RTSP/RTP – if WM uses IETF specs it is only RTSP and not in any actually interoperable way. QT supports pretty much all the codecs, and packetisation formats defined by RTP. And of course it supports multicast – I’m surprised you are unaware of this when 5 seconds googling would find it, Richard. Having spent 2 days last week livecasting conferences with free software, the contents of my backpack and a reflector in Japan, I can confirm that it still works fine.
As for George’s question, which I answered for him then, it is the perverse incentive of monetising congestion I object to. It’s great that people are paying attention to handling congestion in legacy networks, but treating it as a profit centre creates an incentive to congest further, rather than to traverse that inflection point to a higher bandwidth service.
Kevin, Apple products are known for many things, but “wide deployment” isn’t one of them. And I’m glad to hear Apple supports multicast now, I hadn’t noticed since I tend to use the standard products more (PCs running Linux and Windows.) But the real issue here is actually the Internet’s traffic management system.
Those of you who want to turn network engineering and regulation into an Oedipal drama where the bad old phone companies represent your fathers and the brash young startups like Google are your alter egos are missing the fundamental changes that have already taken place in the Internet’s traffic load. In the good old days when the Internet was a private club for elite Universities and defense contractors, the end-to-end model was viable, just barely.
The overloaded Internet of the mid-80s got a fresh breath of air from exponential backoff and slow start in TCP, but only because the most aggressive consumer of bandwidth was ftp and the files it transferred were short.
Now that the Internet has to contend with multi-gigabyte file transfers with BitTorrent, that model no longer works at all. When BitTorrent is slowed down by backoff, it simply propagates more paths, creating more and more congestion. In another year, the Internet is going to be just as unstable as it was in 1985.
This being the case, the carriers have to implement traffic limits inside the network, building on the mechanisms established as far back as the 1980s with RED and its progeny. This is the only way to control BitTorrent.
And while they’re doing that, it makes perfect economic and technical sense to implement voice- and video-oriented QoS. Even Tim Berners-Lee acknowledges this. So whether the phone company manages its links or not, and whether they offer third-party billing for QoS or not, and whether the phone company competes with Akamai by offering content caching or not, the Internet will either change or collapse.
Unfortunately, those of you who have taken up the “neutrality” cause out of your misguided concern about the Internet’s First Amendment are blind to the real issue.
Lots of things support multicast, I don’t think that necessarily was the point.
The point was most providers infrastructures do not support multicast, or at least robust, wide scale deployments. The Mbone doesn’t count.
I would tend to agree with Richard in his view about apple not having a “wide deployment” policy on their products. That said I do believe that with all the “i” products like IPod and such it is set to increase in the near future.