Technology Losers

Tim Bray has an amusing look at seven technologies that seemed spectacular in their day, but turned out to be duds. Here's my thoughts on his list:

• OODBMS. Around 1996, I spent two years working on ObjectStore, which was one of the most sophisticated OODBMS products at the time. I ported the code to various flavors of unix: IRIX, Solaris/x86, LynxOS, etc. Porting can get boring pretty quickly, and the main thing I learned over those two years was how to cope with burnout. During this time, I watched the company go public, and start hemorrhaging money and people. At one point it seemed there were about three people quitting per week. And these were just the people who bothered to send email to all -- otherwise I'd never have known they were gone.
  
• 4GL. I'm still keeping the dream alive, as evidenced by my Declarative Semantics and Dataflow Languages posts. But I've worked with PowerBuilder, and early Visual Basic, and MFC wizards, and so I have an idea as to how difficult it is to write a declarative system that scales well. And that's why I haven't actually released any code yet... ;)
  
• AI. Never been there, never done that. I work exclusively with genuine stupidity.
  
• VRML. I played with it some, but never really got excited about it.
  
• Interactive TV. I am so glad I never worked on that. When I was interviewing for my current job, one of my choices was to work on an interactive TV system. I would have been designing a user interface for scheduling commercials in various time blocks. I wouldn't even have considered it except that it was 2001 and there weren't many options to choose from at the time.
  
• Ada. The first company I ever worked for was founded to write an Ada compiler. They had some success in that regard, but eventually had to find other work. Once, we got a feature request to make the compiler handle a chip with different endian-ness, and I was asked to review a change to the code that implemented some lowlevel primitive, like memcpy or something. It was about 100 lines of assembler, and the diff was just one line. As I recall, the code had to handle about five different special cases, and the change handled them all correctly, but each one for a different reason. That was when I realized at a gut level that assembler programming could be extremely powerful and subtle, and that more experienced programmers had skills I could only begin to imagine.
  
• SGML. Never learned it. I resent it though, because I hate DTD syntax.