Are We There Yet?

The Winter Of Their Content

The last few weeks have seen a few notable Events (please note the capital 'E'). I've been anticipating one of them eagerly, dreading another, and caught completely off guard by the third. The first Event, the one I was looking forward to, was the Transmeta announcement. By now, even if you are Robinson Crusoe, you've heard all about that one. The second Event was the announcement of the super-huge, mondo-giganti-merger of AOL and Time-Warner. I suspect you are just as sick of hearing analysis of that one as I am, but bear with me...I only have a few things to say about it. The third Event doesn't seem to have attracted as much mainstream attention. Not surprisingly, only one of the companies involved has really been talking about it. I'm referring to the patent jackpot struck by Geoworks.

Other than being endlessly hashed over on Slashdot, what can those three events possibly have in common? Before answering that, I have to digress for a little while.

Last year, Harvard Law Professor Lawrence Lessig published a remarkable and truly alarming book entitled, "Code and Other Laws of Cyberspace". I can't possibly do this book justice in the space of this column, but I want to bring up a key argument. In the book, Lessig skewers a sacred cow of the Internet which claims that the net is essentially impossible to control. Lessig points out that this common meme only reflects the current implementation of the network protocols, not an essential quality shared by all communication networks. Another, equally possible implementation could enforce complete control, at the inescapable level of the architecture. In cyberspace, Lessig argues, architecture provides the "laws of physics" that absolutely constrain what is possible and what is impossible. The code is law.

From this perspective, it is easy to understand why the architecture of the Internet today is so difficult to control. From the beginning, it was imbued with the values of its creators: openness, academic freedom, impossibility of subversion. These architectural values embody the constitutional values held by most people in free and democratic states. Other states, however, have distinctly different values. As these states have found that like plugging leaks in a dam, attempts to impose controls that the architecture does not support are doomed. (Following Lessig's lead, I am using architecture in this column as an umbrella term for the transport layer and application layer protocols.) But what about the future? What architectures will the networks of the future have? What values will these future networks make real in their code? If we postulate that the architecture of a space embodies the values held by its creators, then it pays to look at those creators.

First, let's consider AOL/Time-Warner. No matter how much the media companies tell me that the merger is about content, I just don't buy it. It's natural for the media to focus on the media content.  After all it's their business.  But, does Time-Warner really need the extra 20 million AOL subscribers as a content channel? Let's be serious. The typical worldwide audience for a major motion picture measures in the hundreds of millions. Music releases have smaller volumes, but still enormous. Streaming Time-Warner's media content down to AOL users is not going to show off their properties. Time-Warner does have something that AOL wants, though. Only three months after the FCC declined to force broadband operators to open their cable lines to alternative ISPs, AOL purchased the number 2 broadband provider in the country. This deal is all about the pipes. (There goes that cynical voice again, telling me to watch AOL start opposing open cable access.) It wouldn't even surprise me to see the Time-Warner media properties get spun off in a few years.

We can speculate on the values the AOL-over-cable network would embody by extrapolating the values currently embodied in AOL's own network. Offensive speech is prohibited and is grounds for termination. The regulators of the space have one-way broadcasting to the inhabitants. Connecting to the AOL network is not connecting to the Internet, no matter what the commercials say. A user has no choice but to use the client software provided by AOL. That client software facilitates monitoring and one-way commercial traffic. I am not trying to make value judgments about this architecture, only to point them out for your consideration.

At present, AOL is an opt-in system. Your modem will work equally well with AOL or any ISP. Phone companies provide the transport layer access for most users. The ISP provides application layer access to email, news, web browsing. Because these layers are mostly independent, and because each layer is subject to competition, they tend toward interoperable standards. This need not be the case when the transport layer access provider--Time-Warner--also controls the software layer--AOL. It is not hard to imagine AOL client software, tied to the AOL network via AOL broadband. No interoperability requirements, therefore no opt-out whatsoever.  Once Wall Street analysts realize the level of control and monitoring available to AOL under this deal, expect to see AOL's stock come back out of the mild slump it has been in.

If anybody doesn't believe in the value of controlling the architecture, just look at Geoworks.  Their stock hadn't seen the high side of $10/share since January 1998...until they revealed their patent lurking inside of WAP.  Now they are trading in the 30's, with a huge expected revenue stream independent of any product whatsoever.  They don't have to create anything, build anything, or sell anything to keep making money from this patent.  Of course, they have no intention of stifling innovation, just getting a piece of the action.  A piece of all the action.  Companies with revenues under $1,000,000 per year may be eligible for a fee waiver, if the license is executed prior to 1 July 2000.  (Look at the fine print in note 2 for the equivocal language.)  Folks, a consulting company with five people is probably bringing in a million a year.  Do you think they can spare $20K for a patent license?  Of course not.

The WAP Forum membership application clearly spells out the rights and obligations of the companies involved in the Forum.  Paragraph 5.7 states (regarding the company executing this agreement):

5.7 it will grant on request to all other members of the Company ..., subject to paragraph 5.8, a non-exclusive licence to use any of its Essential IPR on fair, reasonable and non-discriminatory terms and conditions. 

The application defines "Essential IPR" as any technology without which a specification cannot be implemented.  According to the terms of this agreement, Geoworks has acted properly.  (So why does it smell so bad from here?)

Here's what will really happen as a result of this brouhaha.  Other members of the WAP Forum who have their own patent arsenals will announce similar licensing fees.  These fees will, by the way, be waived in the event of a reciprocal licensing arrangement.  We should expect to see a knot of entangling alliances to rival Europe in 1913.  Anyone without a war chest full of patents will either pay up or go home.

The WAP architecture looks open. After all, anyone can download the specifications.  In this case, however, control comes not directly from the architecture, but from the interaction of architecture and laws.  The spec says you have to use this patented technology, and the law says you have to pay the patent holder to use it.

Software developers have long said that any problem can be solved by adding another layer of indirection.  Network protocols have a corollary about layers in the stack.  Until now, however, this has not been true of hardware.  Transmeta changes that.  In the past, any x86 clone has been on the Intel treadmill.  The Intel treadmill is the technical device that Intel wields as the defining force in the x86 market.  Any time Intel's high margin products got threatened by a cloner...BOOM, new instruction set.  All the cloners go back to the drawing board, letting Intel have another six to nine month window without significant competition for their new high margin products.  The six to nine month lag comes from the time it takes to get a design into production silicon.  Intel cannot use this level against Transmeta.  In fact, it would probably backfire.  Transmeta might actually be able to get new instructions implemented via code morphing faster than Intel can get them into silicon. 

For Intel, controlling the architecture no longer means controlling the market.  Someone has built a smarter architecture.  The WAP Forum should learn from Intel, or they just might get "Crusoe"-fied .