I found it on Paul Hudak's home page, which also contains some other interesting (although less relevant to lambda) links: support for generating music in Haskell (expanded here); parsing natural language using types; and Fran (for reactive images) which is related to Pan.
Posted to critiques by andrew cooke on 3/20/01; 12:22:54 AM
The two things that struck me the most when reading both papers were a) the (negative) comments about literate programming from some of the evaluators.
b) how much "cleaner" looking the Haskell solution was compared to the included Ada solution.
The only thing missing from the tech report that I would have like to have seen were actual performance figures. It would be interesting to compare the Haskell solution's performance with respect to the other finished solutions.
I wonder how much the authors annoyed the other people - "too cute for its own good" suggests that the report authors thought the Haskell people were smart-alecs. On the other hand, using higher order functions for the regions does seem perfectly sensible and the results from the grad who learnt Haskell in 8 days were impressive.
About speed - I'd like more details too (and memory use). If you find them in the report, please post. There was a table in which Haskell's speed ("Efficient Execution", Fig 4)) was rated highly (an A, compared to Ada's B-). But why isn't the C++ or Awk entry in that figure?
PS Thanks for the positive comments!