Lippard.Multics 1986-11-16 13:56:01 mst Sun
Subject:  Another database saga
Date:  Friday, 14 November 1986 22:44 mst
From:  Paul Dickson <Dickson>
To:  {mbx >udd>m>jjl>misc>misc}

From RISKS-LIST: Digest,  Friday, 14 November 1986  Volume 4 : Issue 11
------------------------------

Date:  Wed, 12 Nov 86 09:50 EST
From:  "Maj. Doug Hardie" <Hardie@DOCKMASTER.ARPA>
To:  risks@CSL.SRI.COM
Subject: They almost got me! [A motor-vehicle database saga] (Mark Hittinger)

     I had a similar situation in college many years ago.  However, the
associated risks were much different.  The school had a honors program in
humanities that replaced al the undergraduate general requirements with one
two-year course.  Competition to get in the program was stiff.  As I remember
the requirements, you had to have all A's in English etc., plus outstanding
scores on the entrance exams.  Only 1 percent or so of each new class was
selected for this program.  It was a real honor and a big deal was made at our
hign school graduation for those who were accepted.  I graduated from
highschool with 2 D's in English and never expected to be considered for this
program.  However, the day after graduation, I received an invitation which I
accepted immediately.  It was a great program.  However, 4 or so years later,
I was running the school's computer center.  The admissions people asked me to
rewrite their program which selected new students for the humanities program.
Since they paid real money, I took the job.  The original program was written
in machine language, not assembler language.  It had one instruction per card
in numeric form.  That was a common approach in the school.  Since the program
was unintelligible, they provided the old algorithm and the new.  It took a
few hours to get the new program working.  Basically, each student had a card
which contained the necessary information.  All that had to be done was to
compare the various values on the card with the criteria and select only those
that met the criteria.  The admissions people provided a deck that had been
run earlier so it was simple to test the new program by running it and
comparing the outputs.  After doing that, we found the new program selected
one less person than the old.  After extensive analysis, we discovered that
the extra should never have been selected in the first place.  That caused
some consternation in the school as it meant that someone who was not
qualified had taken a valuable slot in the program.  So the immediate question
was how many times could this have occurred?  The analysis indicated that
there was only one possible way to be selected improperly and it required a
specific set of values for some 20 different items (including 2 D's in
English).  That set off a bell, and I went back to my hysterical records and
found my copy of my card from years earlier.  There were at least two who made
it through that filter.

-- Doug