Re: [9fans] Re: the 'science' in computer scienscience

[Dan Cross <cross@math.psu.edu>]

In article <20010612150943.1AD92199E9@mail.cse.psu.edu> you write:
>With programming, constraints that helped.  I inherited
>a course at Princeton from Rob.  The thing I liked the best about
>it was an assignment he gave to write a shell.  The constraint was
>in the size of the shell.  We said what it was supposed to do (no
>job control) and took points off for programs exceeding the size
>limits.  I reviewed the best solutions with the students.

Hmm, imposing constraints on people, thereby forcing them to think
about the problem, (or expecting them to, as Laura put it) seems to
be a common thread.  Aside from Anthony's comment about brainwashing
(2 + 2 = Plan 9?), I suspect that, coupled with examples of good
style, might be the most effective way to teach good thought patterns.

>But when it comes down to it, the real challenge is imparting some
>flavor of your taste to the students by example or by the application
>of grades.

Oh, I don't know about grading based on one's personal taste; that
seems limiting to me.  We want to teach people a perspective, but
more importantly to think about the problems for themselves.  If we
mold them in our own *style* (as opposed to way of thinking about
the problem), don't we potentially cut off whole new directions for
solutions?

On the other hand, if we don't impose some limits on taste, we run
into the risk of people doing blatantly offensive things.  Where's
the middle ground?  The humanities have been struggling with this for
centuries.  We (computer scientists) have existed for a scant 50
years, in comparison.

Jon Bentley gave a good example of this in one of the Programming
Pearls books; he gave an assignment to write something, I don't know
what, exactly, perhaps binary search.  Two students turned in 10
*page* programs.  He had a great quote, something along the lines
of, ``Turning a 5 line program into a 10 page one was just too much
to award a passing grade.''  [sic]

>Some rules help, as long as you can get them to understand
>that the rules are guidelines and not hard.  If you're lucky you can
>do it in such a way that doesn't cramp originality on their part.

Yes, exactly....

>I've also found it useful to review Plan 9 code and Unix code with
>them.  That way they can see different solutions for the same
>problems.  I actually find things like the toy operating system
>more harmful than helpful.  They come away understanding some of
>the concepts, perhaps better than reading them in a book, but
>totally lose the perspective of the size of the problems and
>the necessity to be careful with minutia.

The same is undoubtedly true for compilers.  Still, there's an
excitement factor for an undergraduate to see his or her OS project
run which is lost when studying a larger system.

>All this is very interaction intensive.  It's easy if you're a hired
>gun doing one course and then running back to your real life.  I
>couldn't imagine keeping it up semester after semester.

Yes; it seems that it would be rather grueling.  However, I think
that's the difference between folks who are excited by teaching and
those who are excited about research.  Hmm, perhaps excited is too
weak; passionate would be a better word.

	- Dan C.