Linux For Suits

December 2002



Ends and Means

The carrier and content industries aren't buying the Net's end-to-end argument. Here are a couple open source ideas that might finally win them over.


The Internet has been around for a long time, and the ideas behind it have been around even longer. Many of those ideas were shared with UNIX, which has co-evolved with the Net since 4.2BSD was released in 1983. Once that happened, Peter Salus says in Casting the Net (Addison-Wesley, 1995), "Solid protocols on a flexible operating system at low cost was an unbeatable combination. OSI never had a chance."

OSI was an abstract architecture for open systems that turned out to be no less monolithic, and far less appealing, than any vendor's "complete network solution". Oceans of ink were spilled in the 80s on the subject of OSI. Many vendors signed on to endorse and co-develop it (as the Corporation for Open Systems -- COS) in 1985: but ultimately the OSI model succeeded mostly by burning the concept of a "stack" into everyone's brains.

OSI failed, John Quarterman told Peter Salus, because

What got implemented instead was UNIX and the Net. Together they're still getting implemented -- with Linux playing a leading role. But many of these early lessons still haven't sunk in. In fact, the very nature of the Net remains a mystery to many -- perhaps most -- people who aren't familiar with the seminal ideas behind it. Let's look at three expressions of those ideas.

The first is from End-to-End Arguments in System Design, the 1984 essay by Jerry Saltzer, Dave Clark and David P. Reed. This was the argument that framed the Internet's architecture:

The second is something David Isenberg wrote in The Rise of the Stupid Network, the internal document that cost David his job at AT&T in 1997:

The third is from the preamble to The Cluetrain Manifesto, which David Weinberger, Chris Locke, Rick Levine and I wrote in early 1999:

After Cluetrain turned into a book and hit the streets in early 2000, we heard back from a surprising bunch of companies that seemed to be getting real smart, real fast: GE, Procter & Gamble, Ericsson, Wal-Mart, Prudential and Johnson & Johnson, to name a few. Unfortunately for the Net, they were the wrong companies.

The ones that most desperately needed a few clues were the big carriers whose job was to keep things stupid between an exploding number of smart ends. But Verizon, AT&T, Qwest, Time Warner Cable, Comcast, Cox and SBC are all still possessed by their old telephone and cable company mentalities. They still dream about metered traffic and subscription valves on content spigots.

This founding dream so upset Gordon Bell that he circulated an email in January, 1995 with the title "Building Cyberspace with One-way Streets -- bad idea? conspiracy? short-sightedness? incompetence?" In it he pleaded with his readers to put their "bodies in front of the backhoes that are installing asymmetric networks that simply mimic cable TV."

The big "content" companies from the entertainment industry shared those dreams. Unfortunately, they did get the Net's founding clues -- and hated them. The last thing they wanted to see was billions of dumb consumers turned into smart customers -- or worse, smart suppliers. Or, worse than that, both. Which is exactly what Napster showed them. That made Napster very unlucky.

But in the end-to-end ecology of the Net, Napster was just one big mutant animal. It wasn't a species, and it wasn't a habitat. That's why peer-to-peer (P2P) MP3 sharing didn't go away after the RIAA put Napster through the chipper. It couldn't go away, because the RIAA's problem was the Net itself. It defined a vastly social space where everybody could share anything with anybody. Like the assistant's broom in The Sorcerer's Apprentice, Napster's chopped up remains came to life as an army of clones, plus a brand new Internet protocol: Gnutella.

As of September, Slyck.com tallied 2,525,791 FastTrack (a proprietary Gnutella-like service) users online. Slyck say that's 25% more than Napster had at its peak. iMesh has 836,385. Gnutella has 130,556. Slyck also lists six Napster clones, plus 191 MP3 file servers on five networks that range from 3.4 million to 8.3 million songs per server.

The RIAA has taken some of these guys to court, of course, but even if the RIAA succeeds, they can't stop resourceful companies from taking advantage of the Net's end-to-end magic. They can lobby to outlaw certain kinds of businesses, or to regulate them out of existence, as they did with Internet radio in the U.S. But the Net is a worldwide phenomenon. The U.S. feds can't throttle the whole thing. When I asked Dean Landsman, a veteran broadcaster and programming consultant, for his view on the matter, he said this:

I agree. In the long run (and it might be very long), the Net will win. The question for the carrier and content companies will then be no different than it was in the first place: How can I win here?

David Isenberg sets up the answer with a few clues from his original Stupid Network paper:

The "network services" David talks about are ones owned by the carriers, integrated into the network in a highly proprietary way, and sold as value-adds to customers. They are not part of the Internet's infrastructure, which would not welcome them

The Internet, on the other hand, is comprised of many "stupid" protocols that each provide straightforward services. Mail routing. DNS lookup. Hypertext. File transport. Among those, thanks to Gnutella, we now have peer-to-peer file sharing as well. In fact, Gnutella is so peer-to-peer that crossloading seems a better term than of "uploading" and "downloading". Crossloading is a transport concept plainly implied by TCP/IP but not implemented until Napster raised the possibility. In other words, Napster, in spite of its centralized design, did a fine job of making the end-to-end argument.

This argument is so deeply anathema to both the carrier and the content industries that it's success at taking advantage of both is an irony of the first order. So is the fact that the Net withstands continued dysunderstanding by the Hollywood types who desperately want to regulate it away. Witness the paranoia in a list of panel topics from the most recent Digital Hollywood conference:

Thanks to years of constant lobbying, Congress now shares Hollywood's understanding of the Net, and the same paranoia. Tom Poe, who runs a free recording studio in Reno, recently sent a suggestion to Senator Harry Reid of Nevada. Here is the opening to Senator Reid's reply:

Note the language. Senator Reid sees the Net as a space comprised primarily of "Intellectual property" that must be "protected". Stuff made unimportant by lack of ownership is "public domain content" to which "ordinary Internet users" should be "allowed" to have "access". To gauge the ubiquity of this default understanding in Congress, consider this: the DMCA was passed by a unanimous vote.

It's going to take a lot of counter-lobbying to turn that ship around. Four years ago, the authors of the original end-to-end paper (Clark, Saltzer and Reed) were already defending their argument from those who failed to understand it. In that follow-up paper, Active Networking and End-To-End Arguments, they said this:

In his keynote speech at the O'Reilly Open Source Convention last summer, Lawrence Lesssig made a pessimistic case against trying to reason with people like Senator Reid, by quoting departing Congressman J.C. Watts, who said "If you are explaining, you are losing."

Arguing end-to-end clearly works with technologists, but not with most businessfolk, and certainly not with Hollywood or Congress. So that's not what's going to do the job here.

What we need is implementation. Take away Hollywood, the carriers and the feds and you're left with the two populations:

  1. The open source (OSS) developer community; and
  2. Commercial entities that get at least some of the clues.

And, of course, combinations of the two.

We need both these groups, working together, because --

  1. OSS developers are kinds of people who invented the Net in the first place, and who intuitively understand the nature of the Net, which is to support commercial activity rather than to depend on it.
  2. Implementation won't happen unless it eventually supports a market, which is where businesses go to make money

These two last points are chicken & egg problems for anybody who wants to add a new service to the Internet.

Take instant messaging (IM), for example. The Net's founding hackers didn't come up with an IM protocol, but Jeremie Miller and the guys at Jabber.org did, starting in '99. Jabber is very widely used now. IBM and Earthlink both use it extensively on an internal basis. So do uncounted other organizations. Meanwhile, Microsoft, AOL, Yahoo and Apple all continue to lock their users into closed IM systems that don't interoperate with each other. But there's nothing stupid about any of them. Until one or more of them gets stupid, we won't see ubiquitous IM on the Net.

Identity (ID) services is another. Back in June I wrote about what PingID was up to in this new "space". The company (PingID.com) was founded by Andre Durand, who also co-founded Jabber.com. Like Jabber, PingID has a .org counterpart -- PingID.org -- that handles open source development. (Disclaimer: I am on the advisory boards of both Jabber.com and PingID.com.) At the helm of that effort is Bryan Field-Elliott. I recently asked Bryan how it's going. Here's some of what he told me:

I single out IM and ID because I think both can help the content & carrier folks start to make sense of the Net's end-to-end nature. Together those services will make it possible for sellers to have data-rich conversations with highly empowered customers. IM and ID together will support a number and variety of business relationships that verges on the infinite. Those new customers will deliver plenty of clues.

Nothing characterizes the mutant industrial value system we call "consumerism" more than the extreme asymmetry of power between producers and consumers. That value system -- which ought to be called "producerism" -- has been around for so long that we can hardly imagine one comprised of equally powerful vendors and customers.

But that value system is exactly what the end-to-end nature of the Internet was built to support. And there's a name for it, too.

It's called a marketplace.


Doc Searls is Senior Editor of Linux Journal.