Flaming

[utzoo!decvax!ucbvax!menlo70!sri-unix!mo at LBL-UNIX]

From: mo at LBL-UNIX (Mike O'Dell [system])
Well folks, after Bernie's last harangue, I have to put my two cents
in.  Yes, a few well designed libraries would certainly be useful
(regular expressions, etc).  PWB had a feeble attempt and there
may be some such progress in future 4.Xbsd releases.  To get some
idea of what "the second time around" would look like, I strongly
suggest you look at the Software Tools (not the book, but the
software) which were done here at LBL (yes, in Ratfor, but amazingly
portable).  They had the benefit of not having historical baggage
to carry, and were free to steal only good ideas!  The arguments
are more consistant and many such things cleaned up.  It is interesting
to see a "second pass" which was written by the same 3 people
over a short time span.  That for sure encourages uniformity.

On another level, however, I think there is another quite interesting
phenomenon in the evolution of Unix.  Other systems "evolve" at
fairly uniform rates.  Unix as a whole, however, exhibits a very high
instantaneous entropy but really changes (in a global sense) very
slowly.  Part of this comes from the merciful lack of a "development
group" madly gluing frobs on the system to improve its "functionality."
(Such local groups can only screw themselves.)  Another process, however,
is a bit of real "fanatical" dedication to the idea that whatever
goes in EVERYWHERE should be done as right as we know how.  I am
certainly not claiming this always operates (e.g. the multiplexor),
but the reason some things haven't been put in is because noone
knew the "right" way to do it.  Many people knew "a way" to do
something and did it, thus providing the large entropy, but more
importantly, providing the experimental base which eventually
lead to a good solution being discovered.

One other point. I really believe the only way to not get holes in
your stomach is to increase your acceptance of the fundamental
anarchy in the universe (entropy).  Networks provide the best example:
there ain't gonna be no winner and many systems will end up supporting
several different kinds of networks, so spending time fretting
because UUCP builds non-RFC733 headers in counterproductive.  Building
programs to translate between various "standards" is the only
way to keep sane, because no amount of wishing will make the heterogenaity
go away.  Likewise, so you don't like Unix's "cat", well I don't like
the VMS APPEND command.  Do things better in the future if you want,
but realize you are talking about displacing a large mass if you
try to change the past.  

Drivel off.

	-Mike