[Haskell-cafe] Learn Prolog...

[Bill Wood]

On Mon, 2007-09-03 at 02:49 -0400, ajb at spamcop.net wrote:
   . . .
> I'm sure this isn't the case for you, but a typical Prolog programmer's
> idea of "large" is very different from a typical COBOL programmer's.

Ever the diplomat? :-).  Actually that is a fair observation.  I don't
think I ever heard a figure, but I would guess we were in the 20-50 KLOC
range.  The Prolog portion was solving an optimization problem.  The
problem was exacerbated by the fact that the code base was cloned onto
around 20 different machines that were being fed by a 1-to-20 split of a
stream of requests from the internet.  At the back end was FFI access to
a terabyte data store.  Oh by the way, the operational goal was
30-second turn-around from the client end -- submit a request from a
local office and have results 30 seconds later.

   . . .
> Did you look at Mercury?

I looked seriously at Mercury.  It was rejected for two
managerial/political reasons.  The first was that it did not appear that
Mercury could support the scale of the application, at least partially
because it appeared at the time that development was not very active.
The second reason was that the development group had only made the
transition from assembly language to Prolog within the past year or so,
and the prospect of pulling the group through another paradigm shift
made all the managers turn pale.

I sent some email to the Mercury site asking their opinion as to whether
it was up to the challenge; the response was a candid "probably not".

I still sometimes think it might have worked, but the risks would have
been horrendous.

 -- Bill Wood