by Tim O'Reilly
Before getting sidetracked by the Amazon patent furor, I had been meaning to contact you guys about something more immediately important to the Linux community: its lack of a coherent web strategy.
Now, you might say, as I have until recently, that the open source community is doing great on the web front. After all, Apache is the dominant web server; perl, for all its warts, is still key at many web sites; PHP is kicking ASP's butt as an easy to use dynamic page generation environment; Zope is coming on strong as a content management platform.
But there's one fly in the ointment: the lack of a credible client side strategy. This came really clear to me in a conversation that I had at Linux World with an analyst from survey.com. They had done a recent survey of IT managers, and discovered that one of the barriers to entry for Linux in the enterprise was "NETSCAPE IS BUGGY AND SLOW ON LINUX." Given the desire of IT managers to move towards an ASP model for applications, the original Netscape premise, that the browser would replace the desktop as the primary user environment, is coming true...except that it's Microsoft in the catbird seat.
Given this situation, I felt like hitting myself upside the head. I've been talking since my very first public speech on open source in 1997 about how the web, not the desktop, was the critical battleground, but I didn't listen hard enough to what I was saying.
I really see a huge amount of evolution in the web application space, and I see Linux lagging behind. Increasingly, people are writing applications that cross web/desktop boundaries. Many of these applications use features that are specific to Microsoft's browser. Further, Microsoft has made integration of the web into the next generation of Visual Studio a high priority, so that (in theory at least) MS application developers will be building applications that make use of remote web site data as easily as they use local data. Web sites are increasingly thinking about how to create APIs (perhaps xml-based) that will allow application developers to make use of their services other than through the canonical browser. Much of this ferment is leaving Linux behind.
I do see some encouraging signs. I met recently with Andy Hertzfeld of Eazel, and they completely get the web, and are planning to include web features into the next generation Gnome desktop. That gave me some hope. But I keep going back to Mozilla, and the lukewarm support it's received from the Linux community.
I met recently with Mitchell Baker of Mozilla, and she pointed out that the bulk of the 80 Netscape engineers working on Mozilla are doing so on the Windows platform. As a result, the survey.com insight is likely to remain true. Features that were developed for Windows will make assumptions about performance and reliability that won't necessarily hold true on Linux. (E.g. a graphics function that is cheap on Windows might be expensive on Linux.) A simple port may not be the answer.
I think Mozilla needs more resources from the Linux community for testing and optimization, and I'd like all of you to think more seriously about committing them.
But even more than that, I'd like to suggest strongly that leaders of the Linux community get together for a focused technical strategy session with regard to web technologies. Don't just invite people from the linux community, but engage web developers from large web sites based on open source and application developers who are choosing to develop Microsoft-only web applications. What do they need to see in Linux?
Let me know what you think. I'd be glad to set up and host such a meeting if you'd be interested in sending people.