I also made the exception for "small domain" languages. Languages that are suited to do a specific task very well. Examples of "small domain" languages are TeX, sed, sed, gnuplot, but also "languages inside languages", like printf formats, pack formats.

Maybe it's just the way I work, but generally when I need to solve a problem, I first think up an algorithm, and a general outline of the program. Only then I write a program. It will mostly be in Perl (because that's the language I like), but it will lead to a similar program as the one I would have written in, for instance, C. Details will certainly differ, but not the approach.

Abigail

This depends on how "similar" the languages are. Learning Pascal after you've learned C gives you nothing. Learning Prolog or ML/Haskel/Miranda/... can open a whole new world to you. (Though you'll not necessarily like it ;-)

  Jenda

If anything, the Sapir-Whorf hyposthesis is simply backwards: What and how we think shapes our language. The Innuit may have more words for snow (but nowhere near the 40 or so claimed by some reports, more like 7 to 10), but not more concepts or recognitions of different kinds of snow. We just require more descriptive phrases where their language uses simple or compound words. When differences in snow types are a more important part of daily life, one might expect such huffman-like encodings to be reflected in the language.

It seems to me that theories of the spoken word have little value in the arena of programming languages. Language theory assumes that the speaker and the listener are on the same level: they both speak the same language. Programming languages are advanced constructs translated into the language of a very fast, restricted yet capable idiot.

()-()
 \"/
  `                                                     

Not at all. Just because my program is translated into a machine language to run does not mean the machine is the only listener, nor even the primary listener, nor even a listener at all in any reasonable application of the word. The listeners of programming languages are humans. To think otherwise is to fail to understand why we would bother to create higher languages above the machine level in the first place.

You're wrong about the Sapir-Whorf hypothesis, at least inasmuch as you imply that it is so generally considered disproved. This is still an active area of debate and experimental research in some lingustic circles (specifically, those focusing on "embodied" cognitive linguistics, a.k.a. "West coast" cog. ling.), cf. Language and Thought (D. Slobin).

Perhaps I was a little broad with my pen. The SW hypothesis is generally considered refuted in its strong form. Its weak form continues to be investigated and does raise interesting questions in the cognitive sciences. But the weak form is really nothing at all like the strong form of the hypothesis, the weak form merely posits an interaction between language and thought, the strong form asserts that language is a limiting factor in the interaction.