FP: Fairy-tale Programming?
Taking a break from my super top-secret August stealth project, I came across a Joel Spolksy piece on FP, "Can Your Programming Language Do This?". It misses the point, showing "functions as values" as the inspiration for the map/reduce paradigm. He doesn't even show an interesting second-order function -- input only. At least he admits that he's doing nothing you couldn't do in C.
The good thing about Spolsky's article is that it links to Steve Yegge's "Execution in the Kingdom of Nouns". It's a fairy-tale-style explanation of some of the problems Java has: the monomaniacal focus on objects and the absence of "verbs". Clever and very nicely written; the comments are the icing on the cake.
But I still feel like he's missing something. The reason Java is so noun focused is that, as several comments point out, nouns have accountability where verbs do not. I assume that somewhere, someone is very interested in every step their program makes, and they naturally cannot abide an anonymous function. When asked "who is responsible for mapping over that list?", the program can answer this or thatObject. This aptly named "architecture" is necessary for that level of accountability. (That or drastically improved stack tracing, profiling, and live debugging support. When does MzTake get integrated with DrScheme?)
Yegge is right for accusing Java for being in love with architecture, and I think that's the underlying problem: not everyone needs that sort of insanely detailed accountability, but everyone likes to think that they do. Just as I would scoff if people told me that anonymous functions aren't accountable -- I can see them in my IDE during stack traces, what more do I need? -- an 'enterprise' Java programmer must be fairly irritated when told that all that structure is baroque.