[Haskell-cafe] Small flexible projects a possible niche for Haskell - your statement, please...

[Paul Johnson]

On 16/07/10 05:41, Nick Rudnick wrote:
>
>
> In consequence, an 8-student-project with two B.Sc. theses is raised 
> as a pilot to examine the possibilities of using Haskell in the 
> combination small team with limited resources and experience in a 
> startup setting - we want to find out whether Haskell can be an offer 
> competitive whith languages like Ruby & Co. in such a setting.
>

I'm not sure exactly what you are asking, but I'm going to try to answer 
the question "Does Haskell have a niche in small, flexible projects?"

I think the answer is a definite yes.  I also think that Haskell can do 
great things in bigger projects as well, but successful technologies 
often start out with a niche that was previously poorly served, and then 
move out from there.

Haskell developers generally start by writing down an axiomatic 
definition of the problem domain.  To a developer raised in traditional 
"top down" development this looks like a jump into coding, and 
furthermore coding at the lowest level.  In fact it is a foundation step 
in the architecture, because Haskell works well with a "bottom up" 
approach.  The property that makes this work is "composability", which 
says that you can take primitive elements and integrate them into bigger 
units without having to worry about mutual compatibility.  A Haskell 
library will typically define a data type "Foo" and then have functions 
with types along the lines of "mungFoo :: Something -> Foo -> Foo".  
This "combinator" style of library give you the
basic building blocks for manipulating Foos, along with a guarantee that 
the output will always be a valid Foo.  So you can build up your own 
applications that work at the "Foo" level rather than down in the coding 
level of flow control and updated variables like conventional programs.  
This lets domain experts read and comment on the code, which reduces 
defect rates a lot.

But these combinator libraries are also highly reusable because they 
describe an entire domain rather than just being designed to fit a 
single application.  So the best bet is to analyse a domain, write a 
combinator library that models the domain, and then produce a series of 
related programs for specific applications within that domain.  That 
will let a small team be amazingly productive.

Paul.