"But comparing a C compiler and a Haskell compiler is like comparing Babe Ruth to Pele." I was thinking more along the lines of a chain saw and an Intergalactic Telefunctor. ;)

"I don't know how popular computer science has adopted such a nonscientific culture." I'm willing to bet the marketing department had something to do with it. Just a hunch.

It would be useful to have a bona-fide economic analysis of the costs/benefits associated with the two styles of programming, but you'd have to have a combination economist/computer scientist to do it. For example, testing and bug fixing has a direct cost, but also an indirect cost of lost opportunity - time spent hunting down and fixing bugs could be spent improving the product. It would be interesting to see real (or at least principaled) numbers. I suppose eventually, once the real money starts flowing towards FPLs, we'll see something like that.

Sorry about the 30 character thing. I'll file a bug report.

Thanks,

-gregg

The "ghc vs. gcc" discussion is silly, and flame wars derived from such a comparison perpetuate harmful patterns of thought. Maybe it is possible to compare two C compilers objectively and say that one is better than the other, but you'd need a lot of evidence. But comparing a C compiler and a Haskell compiler is like comparing Babe Ruth to Pele.

gcc when given input A produced code which ran in less time than ghc when given input B. But we have no theory telling us that A and B are the same program; simply that they print the same number when run. But I can write C (and Haskell) programs which are much faster and much slower than that benchmark, which print the same number.

This difference isn't purely academic, however. I bet that I can write a lambda calculus interpreter in Haskell which a C programmer would be hard-pressed to compete with. People want fast lambda calculus interpreters; such cannot be said for programs which print out some arbitrary fixed number.

I hate language shootouts, benchmarks between languages. Speed is not a property of a language, it is a property of a program. Unless you can show that two programs are the same (intensionally!), comparisons are meaningless. And even if you do show that they are the same, your comparison has correctly compared the speed of the compiler for one input.

I don't know how popular computer science has adopted such a nonscientific culture.

 </rant> 

(Oh, also, whose bright idea was it to restrict the website field in your blog to 30 characters? :-)