Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!ccplumb
From: ccplumb@watmath.UUCP
Newsgroups: comp.edu
Subject: Re: Computer Science: where it belongs
Message-ID: <15484@watmath.waterloo.edu>
Date: Wed, 11-Nov-87 02:19:37 EST
Article-I.D.: watmath.15484
Posted: Wed Nov 11 02:19:37 1987
Date-Received: Fri, 13-Nov-87 05:57:13 EST
References: <469@ndsuvax.UUCP> <16118@clyde.ATT.COM> <180@spock.UUCP> <4422@pyr.gatech.EDU> <3355@ames.arpa> <21402@cca.CCA.COM>
Reply-To: ccplumb@watmath.waterloo.edu (Colin Plumb)
Organization: U. of Waterloo, Ontario
Lines: 48
Summary: This discussion would interest RMS

In article <21402@cca.CCA.COM> g-rh@CCA.CCA.COM.UUCP (Richard Harter) writes:
>[SE's should spend more time learning from other programs.  Two points:]
>
>One is having a bank of pre-existing program plans to draw on.  Doesn't this
>make sense?  Architects don't sit down and design bridges and buildings as
>though no one had ever built a bridge or a building before.  Suppose you have
>to design a system.  Wouldn't it be sensible if, for each requirement, you
>could go to the design catalog and get a list of system features that would
>be needed.  Wouldn't it be sensible if you had a suite of major architectures
>to select from instead of designing one de novo?

Yes, it would.  But where do you get it?  Richard Stallman has flamed
many times about the needless duplication of effort engendered by the
common impossibility or illegality of using another program's source as
the basis for some product.  The GNU project should improve the
situation considerably.

>The other point is that SE people should spend some time studying large
>programs.  What is the principal decomposition?  It's all very well to be
>able to write small programs that implement neat algorithms, but a lot of
>people are going to have to work with large systems.  Shouldn't an SE graduate
>have some experience in finding her way around a large system?

Again, most large systems are developed by people who wish to make
money licencing their use.  They would tend to object to having several
hundred SE students per year investigate the innards of their
proprietary code.

There are, of course, notable exceptions.  GNU Emacs, the netnews
software, TeX, and X-windows come to mind.  Testing several hundred
bug-ridden modified versions of the news handling software would
require the investment of considerable effort in the construction of a
realistic playpen if live testing, and the resultant chaos, is to be
avoided.

The others, however, would work nicely.  The only problem would be
providing the resources to handle a copy of the sources per student,
and many, many compiles per student.

Of course, one could aim a little lower with Mush or MH, or Nethack
(sure to receive extremely thorough testing).  If anyone has a course
based on something similar to this, I'm sure the net would be interested.
--
	-Colin (watmath!ccplumb)

Take *everything* with a grain of salt dept.:
"America could gain much from Russia's long experience with Mars."
	- The Economist, Oct. 3.