Re: Request for information
Jack,
I think it will be tough to sell Lisp over more popular languages such as
Java or C++.
One strategy is not to mention language-specific features at all. Instead,
I think you should talk about the costs and benefits of switching
environments. For example, you could say that Lisp is a higher-order
language than C++ or Java, and that switching would result in a much larger
system that would require more manpower and higher maintenance costs. You
could back that up by with this paper:
==================================================
"Using Common Lisp to Model and Validate Mission Critical Software"
David Lamkins
Proceedings of the "Lisp in the Mainstream - 40th Anniversary of Lisp"
Conference, 1998 -
Abstract:
"At DRC [Dynamic Research Corporation], we used Common Lisp to build an
independent Model of a large C++ system, then compared the production
system's operation to that of the model. This exposed several latent flaws
in the production code, which we were able to correct before delivering the
system to our customer. The simplicity and compactness of the Lisp model
was an aid to understanding and validating system operation. Two thousand
lines of Common Lisp code captured the essential features of 30-40 thousand
lines of C++ code. The model's throughput was initially within a fractionO
of three of the production system when executed on identical platforms. At
the end of the eight-week project, one simple optimization reduced the
performance difference to less than ten percent. Thanks to our positive
experiences, we anticipate additional uses of the model code as: a testbed
for new feature development, a portable sales tool deployed on laptop
computers, and a core around which a low cost product can be developed for
a new target market."
==================================================
This paper is relevant because it deals with a telecommunications fraud
model. (Sorry, I don't know where to find this paper online, but you should
contact the author).
Then maybe you could talk about the problems and expense in rewriting a
working lisp application in another language. You could say that re-writing
would not be a sure bet. You could refer to this anecdotal thread where
Paul Graham talks about the re-writing of Yahoo! Stores from Lisp to C.
http://www.ai.mit.edu/~gregs/ll1-discuss-archive-html/msg02367.html
I also remember reading an article on comp.lang.lisp where an engineer at
Parametric Technologies describes a product written in 3500 lines of lisp;
their competitor had advertised an equivalent product written in 900,000
lines of C++. I can't seem to find it now with google groups - it may have
been X-archived.
I wish I could help more; I think talking about higher expenses and possible
deterioration in the quality of service is the way to go.
----- Original Message -----
From: "Jack Gray" <jack.gray@aexp.com>
To: <lisp-hug@xanalys.com>
Sent: Thursday, April 17, 2003 1:59 AM
Subject: Request for information
> Hello everyone.
>
> I have worked for American Express on the Authorizer Assistant (AA)
project since its inception
> in the mid-1980s. I've been told that AA is generally recognized as one of
the oldest and most successful commercial
> LISP-based applications - and I personally believe that the prime reason
for that success is LISP.
>
> The AA project staff has been asked to reassess AA from top to bottom,
with an eye toward developing a future
> direction for the application. (As you can probably guess, there will be
pressure for us to consider a re-write
> in some environment other than LISP.)