(Excerpts from a message dated 11-03-99, Ray Hyder to Murray Lesser,
original topic: OS2 2.1 IS Dead):

Hi Ray--

 ML> bugs in the programming interface (IMO, more bugs have been 
 ML> introduced by FixPaks than were in the original code).  Early 
 ML> CSDs did not add new features. 

RH>Murray, my experience is that the newer fixpaks are well tested by
  >IBM and introduce very few, if any, problems.  I'm at OS/2 V4 with
  >FP12 installed. It is the best version of OS/2 I've seen yet.  Very
  >smoooth and reliable. 

    As any programmer knows, it is impossible to test complex software
to the point that one can guarantee that it is bug free.  (A very old
programmer's adage is that the only bug-free software is software that
is no longer in use!)  This is particularly true for an operating system
that can be installed in many different configurations on many makes
(and combinations of makes) of hardware.  You might take the time some
day to check out how many of the new files in your newest FixPak replace
those same-named files that were updated in a previous FixPak.

    My experience has not been the same as yours.  I have had too many
FixPaks that broke programs I had been using for years before, and I had
to back out of them.  My most recent example was FixPak 6 for Warp 4,
which failed to boot at times :-(.  (Such intermittent bugs are very
hardware dependent.)  So I backed out to Warp 4 FixPak 5, which I am
still running.  I'm not going to try later FixPaks just to see if that
bug was fixed!

    Also, some HPFS-related FixPak since the last time I formatted an
Iomega Zip diskette HPFS (under Warp 3 FixPak 5) has modified my system
so it can no longer do that.  I learned about this when trying to format
a Zip diskette HPFS the other day as a test before attempting a reply to
a query asking how to do it.  Since I had run enough tests under Warp 3
to determine that I had no personal interest in Zip diskettes formatted
HPFS, I haven't attempted to chase down this newly discovered
FixPak-induced bug :-).

    Not to speak of the high probability of the FixPak writers making
poorer design decisions than were made for the "unfixed" original, so
their FixPak has to be fixed again with later FixPaks :-(.  For example:
Consider the "Y2K fix" in Warp 4 FixPak 5, for the file date returned by
the REXX STREAM(...QUERY DATETIME) function.  It was easier to program
the correct four-digit year for all allowable file dates from the
original two- or three-digit "buggy" return than it is to derive the
correct four-digit year from the two-digit values always returned by the
"fixed" version.  Furthermore, It isn't possible to derive the correct
four-digit year from the values returned by the "fixed" version for file
dates after December 31, 2079--if anyone cares!)  To make matters worse
for the REXX programmer (and that program's user) these two workarounds
are incompatible, so REXX programs that use the "wrong" workaround for
the REXX interpreter level won't produce the correct answer.  (There is
a workaround for this, but you get the idea.)

    To add insult to this injury, the FixPak gremlins introduced two new
fixes into FixPak 6 for the problem they had created in FixPak 5: the
all-new STREAM(...QUERY TIMESTAMP) option and a new switch for the
SysFileTree() function, to return file dates with four-digit years.  Of
course, if you use either of these new function forms in your REXX
program, that usage will throw a "syntax" condition when anyone attempts
to run that program under pre-FixPak-6 REXX interpreters.  So now you
have to be careful about whom you give it to :-(.

    I am sure that I could find other FixPak misadventures than those I
have related here, if I tried very hard.

 ML> * MR/2 2.25 #120 * If it ain't broke, don't FixPak it. 
 
RH>Murray, I used to feel the same way.  "If it ain't broke don't fix
  >it!"
 
RH>My advice now would be to get to the current version of OS/2.

    Very poor advice.  Have you tested enough to be certain that your
current system is really less buggy than it was before you applied your
latest FixPak?

    After over six years of running various versions of OS/2, and too
much experience with ills introduced by FixPaks, I have learned not to
apply a FixPak (after the first few following the introduction of a new
version of the operating system) unless it supports new hardware that I
intend to install, or it adds new function that I want.  I have
carefully read the README for Warp 4 FixPak 10 (the most recent that I
have on CD-ROM), but have decided not to try installing it because there
is nothing in it that I need.

    De gustibus,

        --Murray

___
 * MR/2 2.25 #120 * If it can happen, it will (Murphy)

--- Maximus/2 2.02