Sad, I know. I had dreams of getting some serious blogging done while I was on vacation, but they all vanished in a haze of roadside attractions and chile rellenos. Instead, you get these tidbits:
- Steve Yegge says “the worst thing that can happen to a code base is size.” I don’t know why he had so much trouble with Eclipse — I’m looking at… find, find, wc, awk… just under 7000 classes and a million lines of code and it seems to work fine for me — but he’s got a point. At my last job, with about half that much code, we were able, just barely, to take our organically grown mess and decompose it into sensible layers and functional units, when a new choice of tools (WebLogic Workshop) forced us to, and it took half a dozen developers a month. If we tried it here, I wouldn’t be too suprised to see it take four.
- In the wake of the Javapolis closure discussions (notes forthcoming — no, really), Bruce Eckel thinks that “Once you bind yourself to backwards compatibility with anything, you must be prepared for your language to be corrupted as you add features. If Java is unwilling to break backwards compatibility, then it is inevitable that it will acquire both unproductive complexity and incomplete implementation of new features.” True dat. But up against it, well — did I mention a million lines of code? It seems like there’s an unsolved problem somewhere in there, about the evolution of large systems. (And no, Ben, switching to a job where everything is a small one-off consulting project isn’t always the answer.🙂 )
- Martin Fowler finds that clients (as in, customers) can’t be trusted to keep the test bar green. Maybe ThoughtWorks ought to write it into their contracts that they’ll charge for the time it takes to get the bar green again before making any enhancements.
- Tim Bray has some thoughts on providing wireless at conferences. Javapolis was my first experience with a fully (rather flakily, especially when two thousand people would all simultaneously leave a session and start checking email, but) wifi-enabled conference, and overall I agree with Bray that it’s a big plus. This isn’t school; your audience doesn’t have to be here, and if you can’t keep them interested, you shouldn’t be speaking. Mostly I wish I’d screwed around more during the conference, not less — maybe I’d have discovered in time to ask James Ward about it that FlexBuilder (contra the Flex docs) doesn’t let you assign to a
constfield in the constructor.