What we really need is a sort of stress-strain curve for each of the major languages. Since Haskell is a typed language, we can have one curve for types and one for values: VARIABLE TYPE VALUE ------------------------------------------------------------------- stress | effort to learn language | coding effort/time required -----------|---------------------------|---------------------------- strain | ability to solve problems | marginal rate that problem | | is being solved -----------|---------------------------|---------------------------- yield | knowledge needed to write | boilerplate code needed strength | "Hello World" program | -----------|---------------------------|---------------------------- modulus of | semantic power of | productivity once boring elasticity | language syntax | stuff has been written -----------|---------------------------|---------------------------- ultimate | expressive power | NONE strain | of the language | -----------|---------------------------|---------------------------- ultimate | NONE | point at which code strength | | is getting beyond you -----------|---------------------------|---------------------------- failure | NONE | point at which code is point | | broken and indecipherable -------------------------------------------------------------------- Each language will be strong in one part of the curve. Haskell is superior in those parts of the curve where it matters most in real tasks, at the high strain end of the graph, both in type (there's always something more to learn, so no programmer burn-out) and in value (one person can manage more complexity with less effort). The PR problem is that newcomers to Haskell are being asked either to: 1) Trust me (but President Bush has strained that argument past failure) 2) Sample the curve at the low end (benchmarks, toy problems) and extrapolate the higher end, giving a very false impression The only answer is to provide a positive marginal interest at each point in the language acquisition process to entice the learner to keep sampling as (s)he progresses individually up the curve. This is the real benefit (and most noble purpose) of haskell-cafe. And of course the justification for this strained material science metaphor! :) Dan Weston Philippa Cowderoy wrote:
On Wed, 10 Oct 2007, Andrew Coppin wrote:
(I'm less sold on whether you really need to learn a particular dialect well enough to *program* in it...)
If you don't then you won't be able to see how complicated things actually get done. It's also an important exercise in abstracting things and keeping something understandable when the system you're building is fighting back against it.