Saturday, July 23, 2005

Michael Tiemann on Open Source

The peerless features a talk by Michael Tiemann of RedHat, formerly of Cygnus and the guy who first wrote the GNU C++ compiler. Some interesting points from the talk:
  • In 1995, when MySql started (this was a talk at the 10th anniversary MySql conference), Tiemann had already turned over the reins of the company to a president at Cygnus; the company was then making $6 million in revenue and employed more than sixty people, disproving the fact that you cannot do business with open source. Larry McVoy was kicked out of Sun in the same year for advocating open source as a strategy.
  • McVoy brought to Tiemann's notice a small company in North Carolina called Redhat - Tiemann advice to acquire a stake in that company was ignored and five years later, Redhat bought Cygnus.
  • It took him two years to figure out how to do business with open source.
  • Tiemann addressed the question of why OSS had succeeded and drew an analogy with the observation made by Alexis de Tocqueville in 1830s, that the sum of all individual undertakings in America exceeded the efforts of the government.
  • Disruptive technology always comes from unexpected quarters.
  • Cost cutting is not always the most interesting question - there is always more upside to inreasing revenue than to cutting costs. There are limits to cost cutting, while there are none to increasing revenues.
  • The real value of architecture is when something comes along and it can be accomodated quickly. Explained the value of strategy with an example from a Wall St. firm - a strategy that was 10-20% different resulted in a ten-fold increase in growth.
  • According to research done at CMU, the Apache project has ten to fifteen people who do eighty five percent of the work. But the total number of people who contribute is closer to 400 ( 388 ?). It's this extra non-core contributions that add lot of polish and quality to the software. Tiemann asked a rhetorical question about going to a venture capitalist to ask for money to recruit all 400 developers. (I thought this was a particularly effective illustration of the role of a large pool of contributors - more about this aspect in a future post, if time permits)
  • Quoted Eric Von Hippel, faculty at MIT and the author of Democratizing Innovation, on the role of user participation in design. EVH in his book explains in his book how Jack Welch managed to make GE a leader in plastics by bringing in the concept of user design toolkits. Previously GE's business model made it possible to get into selected market segments, in fact the largest customers. Jack Welch said that GE doesn't need to be the exclusive designer of plastics - Why can't customers design their own plastics ? This idea of moving of moving the locus of design from the supplier to the customer resulted in 85% of the designs coming from the customers ultimately. MT then asked "What rational person would let go of this opportunity ?"
  • Drew some analogies between disruptive open source technologies and themes from the book "Guns, Germs, and Steel" by Jared Steel. (The analogies about the conquest of the native american population by the Spanish seemed out of place to me)
  • He recalled reading Stallman's code for the first time - first time he said, he saw good code. The architecture and implementation unfolded in one view.
  • He remarked that even the newly elected board of the OSI did not have a clue about the reasons for the success of open source. That's the territory that needs to be protected and grown. He urged the audience and sophisticated users to think about it.
  • Over 70% of all projects on are covered by the GPL or BSD style license.
  • Quoted designer Bruce Mau about the fact that design is not very visible unless there is a failure and used Micrsoft's out-of-control spending on security to conclude that the shared shared model was a design failure. Microsoft's spending on security in Jul 2002 was $100 million and in Mar'2005 $2 billion. At this rate they should end up spending the rest of their bank balance on security. MT contrasted this with the improvements in open source software. The Fuzz report noted a 20% failure rate in Unix utilities in 1985. In 1990, by which time the GNU tools had appeared, the failure rate was down to 25% of that of Unix tools. In the same period the Unix utilties/tools had not improved substantially. Later the failure rate of GNU tools went down to 2% and by 1996 the GNU tools were a 100% clean.
  • Software patents are the lead banner for massive stasis - it really means that innovation will stop for the next twenty years.
  • Referred to Bruce Mau's work which rejects the notion of a client-designer relationship. The notion of an open source license may be obsolete, when it tries to break the above law.
  • How to enage the plurality of potential contributions in OSS.
A recommended addition to your IPod/whatever. A very thought provoking lecture from a person who has written a compiler, founded a company, been a CTO - a man of many hats.

Update: Looks like Michael Tiemann gave a similar talk (pdf) at the Redhat summit.

No comments: