NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/comp/edu/1531

Internet Message Format | 1992-09-04 | 3.7 KB

Xref: sparky comp.edu:1531 comp.lang.fortran:3426 comp.lang.misc:2932 comp.arch:9223 sci.math:10989
Path: sparky!uunet!decwrl!purdue!mentor.cc.purdue.edu!pop.stat.purdue.edu!hrubin
From: [e-mail] (Herman Rubin)
Newsgroups: comp.edu,comp.lang.fortran,comp.lang.misc,comp.arch,sci.math
Subject: Re: Scientists as Programmers (was Re: Small Language Wanted)
Message-ID: <[e-mail]>
Date: 4 Sep 92 14:20:14 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail] (USENET News)
Organization: Purdue University Statistics Department
Lines: 55

In article <[e-mail]> [e-mail] (Ari Juhani Huttunen) writes:
>In article <[e-mail]> [e-mail] (Raul Izahi Lopez Hernandez) writes:

>!    A good compiler and a good optimizer can help any scientist to write
>! reasonable code, however there is no software yet that can help a CS
>! person write any physics code.

>What do you mean by 'reasonable code'? Do you mean code that solves correctly
>the test problems that were available when the code was written? Or do you
>mean code that solves _any_ problem it was intended to solve, regardless
>how different the actual problem is from the original test data?

>If the former is what you mean by 'reasonable code', then go ahead and
>use untrained programmers and rely on the compiler and optimizer to
>do the job. (NOTE that I did not write 'physicist' in that sentence.)
>If the latter is your concern and you want correct solutions, use
>a professional!

>To produce first quality code, the best solution would in my view
>be to use 1) a physicist 2) a computer science person, paired up in
>a team. Both should know something of the other's profession. Neither
>is sufficient alone. Of course, a person that can do both would be
>better, but being mere mortals, such persons are difficult to find.

This is exactly what we should have been doing all along.  There seems
to be a tendency everywhere to try to either break up a task (like
having one doctor treat the heart, and another the lungs, and never
having them communicate; or having the physicist write the formulas,
and the programmer try to program them, without the programmer having
any idea even of typical values), or to set things up so that a single
individual "should" be able to do it alone, like setting up packages
which SUPPOSEDLY do the statistical procedures which any social
scientist should want to use.  Both of these necessarily fail.

Now most universities provide statistical consulting.  One of the
big problems is to convince the user that it is not enough to dump
the data on the consultants desk (or to give the consultant a tape
or a disk) and expect the proper thing to be done.  There have been
postings on many groups asking to "fit a curve" to data.  Now I do
not know which of the many versions of this a client means without
questioning the client.  The biggest problem, in practice, is that
the client does not know what is wanted.

It is also the case that mathematicians in universities will often
provide help.  But nothing like this is available at the programming
level; there is no problem in getting government support for faculty
to program their own jobs, but not for paying graduate students to
work with them on research which cannot be used in doctoral dissertations.
The consulting facilities provided by computing centers do not extend to
helping with programming.
--
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
Phone: [phone]
[e-mail] (Internet, bitnet)
{purdue,pur-ee}!pop.stat!hrubin(UUCP)