Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!hplabs!hpfcdc!hpldola!hp-lsd!tbc
From: tbc@hp-lsd.HP.COM (Tim Chambers)
Newsgroups: comp.edu
Subject: Re: models (was Which language to teach first?)
Message-ID: <7870002@hp-lsd.HP.COM>
Date: 31 Jul 89 15:52:22 GMT
References: <8514@batcomputer.tn.cornell.edu>
Organization: HP Logic Systems Division - ColoSpgs, CO
Lines: 47

Having learned Lisp as my first language under the instruction of Sussman, I
agree with the points made by manis@grads.cs.ubc.ca.  The language should be
irrelevant.  I won't try to speak about the current MIT environment, but
during my enrollment (1980-84), one was hard-pressed to find any course that
taught a language.  I was immersed in the philosophy that the theoretical
material (data abstraction, automata, algorithm design, etc.)  should be
taught during the class and that programming languages should be given no more
attention by the instructor than was necessary to make it possible to
accomplish the lab work.  We were given a reference manual and sent off to do
the assignments.  Examples during class were limited to discussion about the
concepts, and Lisp syntax was freely used, but minimal attention paid to Lisp
semantics.

I do not understand what Mr. Manis implies by the following, though.

|There is, however, the non-major population, who clearly neither want
|nor need Abelson and Sussman. With this group, programming is clearly
|not the major issue, and Pascal is quite suitable.

If programming is *not* the major issue, why force them to study the theories
using Pascal?  If you want to teach abstract concepts, what better vehicle than
a language that requires thinking in terms of a few fundamental concepts --
atoms, lists (ordered sequences), and recursion?  Sussman stressed that we
were using Lisp because it was a simple, yet powerful language to practice the
concepts being taught.  Arrays, structures, vectors, complex numbers -- all
were introduced to us in their abstract form and then we practiced
manipulating the abstract types by representing them with lists.

I learned Pascal on-the-job during a summer internship after learning Lisp and
CLU in school.  The semester I returned, I confronted my software engineering
professor about why I had to learn CLU to do my labs instead of using an
"industrial" language like Pascal or C.  He reinforced the school's philosophy
that precious class time should not be wasted teaching language particulars;
rather, if we had learned a basis for software design, we could apply those
principles to *any* language we might encounter.  (Indeed, CLU was a "pet"
language of the department, but it was *tuned* for teaching good
object-oriented design techniques before Ada or, I believe, C++ and
Objective-C, were available.)  After five years in the "real world" I can
attest to the wisdom of that approach.  I have seen the merits of Pascal,
Gosling's Mock-lisp, C, Ada, Objective-C, C++, Smalltalk, Prolog, and assembly
language debated during my career.  My education allowed me to adapt to
whatever language my employer has required me apply.

Please, teach the concepts *first* with *minimal* attention to the language
representation.  Then introduce languages as *examples* (some better than
others, of course :-) of the concepts of data abstraction and algorithm
design.