COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\input rwflet
C00008 ENDMK
C⊗;
\input rwflet
\memoto Faculty, Ph.D.
\from Robert W. Floyd
\subject Comprehensive Structure; Long-term Committees

\body
My opinion on the comprehensive exam, repeated from lunch of February 14.

Our historical experience with the comp over perhaps fifteen years has until
recently been tranquil; therefore, I think our current dissatisfactions can
not be inherent in having a requirement for passing a breadth exam in two
years.  On the other hand, the exam does not get much use as a final filter,
in that almost everyone eventually passes it, so we might consider
extending the deadline to 2.5 or 3 years, to counterbalance proposed
early research requirements.

The comp embodies the material of a good many courses.  The theory exam
covers much of the material of 160, 154 or 254, 257, 260, and more.  Only
very well prepared entrants can hope to pass the comp after one quarter
of study; historically, no more than our top (perhaps) 20\% have done so.
Students should not be led to feel they have failed if they do not
pass the first time.  In fact, a student does not fail the exam until the end
of the second year.  A non-pass is not a failure.  A reasonable goal is to pass
one section each  half year.

It is difficult to be on the comp committee, because we come onto it rusty by
several years, and find a new syllabus and regulations awaiting us.  The comp
committee, like several other major committees, should be a continuing
body with staggered three-year terms of appointment (like the major university
committees).  This permits the members, also, to maintain collections of 
unused problem ideas, and to calibrate problems by experience.  One virgin per
wedding night is plenty.

The current format of the comp sections, each broken into five named parts
with a tacit assumption of several problems per part, leads to a three
hour exam of a dozen or more problems, at best allowing fifteen minutes per
problem.  This can not encourage students to reveal the depth of their
understanding.  It also unduly contrains the committee to cover all the bases
rather than ask the best available questions.  I would rather present eight
nontrivial questions, with the grade based on the student's five best answers,
to discourage wasting time trying for partial credits.  Uniform coverage of
subdisciplines should not be expected.

I think we should prune subject matter.  Areas subject to rapid technological
change especially should be prayerfully left to the student's self-education
in later life.  Networks might go away, operating systems and compilers 
could be covered
only at the level of a good software engineering survey text.  The ruling question:
is the half-life of this technology greater than the median time to Ph.D.?

I return to a proposal briefly mentioned, namely long-term committee appointments.
Having been a long-term member of the MS committee, I have appreciated reading
MS admissions folders in the light of a set of standards set by experience; it
cuts the hard decisions in half, at least.  Some committees, such as
curriculum, are subject to excessive thrashing when staffed in a short-term
way.  On some committees, ex officio members seem to have a disproportionate
influence because of their longer tenure.

Couldn't we each be on one major committee for a (renewable?) three year term?
The committees I have in mind are Ph.D., MS, Undergrad, Curriculum,
Forum, and an executive council (i.e., the area spokesmen, with much 
authority delegated to them).
\endletter
\end