Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site topaz.ARPA
Path: utzoo!linus!philabs!cmcl2!seismo!columbia!topaz!@RUTGERS.ARPA:Bruce_A._Hamilton.OsbuSouth@Xerox.ARPA
From: @RUTGERS.ARPA:Bruce_A._Hamilton.OsbuSouth@Xerox.ARPA
Newsgroups: net.works
Subject: Re: Not again!?  Assembler vs High-Level languages
Message-ID: <1752@topaz.ARPA>
Date: Wed, 24-Apr-85 00:34:44 EST
Article-I.D.: topaz.1752
Posted: Wed Apr 24 00:34:44 1985
Date-Received: Fri, 26-Apr-85 06:21:43 EST
Sender: 0ð$ÿìPÀA@topaz.ARPA
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 52

From: Hamilton.OsbuSouth@Xerox.ARPA


------------
    From: Doug Bryan <BRYAN@SU-SIERRA.ARPA>

    domains where size and speed are very important...
    are becoming fewer and fewer...readability and
    understandability become the major qualities of
    "good code".  
------------    
    From: Keith Doyle
    {ucbvax,ihnp4,decvax}!trwrb!cadovax!keithd

    Size and performance have no direct correlation.

---------------
This discussion should move to SOFT-ENG.

There's a lot of horsefeathers floating around here.  First, size and
speed will always be very important, except in the most trivial
applications.  After 30 years or so, I hope we've all learned that EVERY
computing resource (like every other human resource) gets used to the
max.  Especially in today's cut-throat marketplace, all of us except
MAYBE Grid and Cray will continue to try to stuff ten people into a VW,
so to speak.

Size and speed "versus" readability and understandability is a strawman.
If code is elegant (i.e. readable and understandable), it will tend to
be efficient in both size and space.

There's no magic in HOL's.  The Xerox Star and Network Services are
coded entirely in Mesa (similar to Modula-2), in the
million-lines-of-code ballpark, supported by a smattering of
special-purpose microcode, and have encountered the whole ball of wax in
performance issues: time vs. space tradeoffs, project management
questions, how to indoctrinate new hires,...  I don't care HOW
high-level the language, questions of individual style arise and local
conventions must be arbitrated, in any large system.

I grant that HOL's give some readability advantage and (with type
checking) prevent a lot of dumb mistakes, but they also make it easy for
the incompetent to generate hundreds of lines of HOL garbage, instead of
dozens of lines of assembler garbage.  Nothing substitutes for talent.

Size and performance sure ARE correlated.  My disk doesn't swap in zero
time.  Even if you will soon be blessed with a personal workstation with
multimegs of RAM that can always hold your entire working set, that only
moves the problem to the database servers, bitmap processors, etc., that
you interact with.

--Bruce