Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!bloom-beacon!husc6!cmcl2!phri!dasys1!tneff
From: tneff@dasys1.UUCP (Tom Neff)
Newsgroups: news.admin
Subject: Re: A counter-example for those who would eliminate X
Message-ID: <5330@dasys1.UUCP>
Date: 5 Jul 88 06:12:15 GMT
References: <264@octopus.UUCP> <3302@s.cc.purdue.edu> <265@octopus.UUCP>
Reply-To: tneff@dasys1.UUCP (Tom Neff)
Organization: Independent Users Guild
Lines: 55

Pete Holzmann points out that humongous machine-specific *source*
postings are appearing on the net recently, and suggests a comparison
with PC binaries.  He insists that he's not disapproving of, for
instance, the Mahjongg tiles, but it seems to me there would be little
grounds for comparison if you couldn't make the case that they, like
the PC binaries, are signs of a Nasty Trend.

And I think he's got a point.  If you're going to post source to
thousands of machines, most of which are not Suns, then you should have
compile-time switches that let the user build versions for other
machines.  If that's not possible because an application's too
hardware-specific, then don't post it to comp.anything -- set up ftp or
mail it to requestors.  The "source is source" excuse is no excuse.
Source can be anything, trash or gold, suitable or unsuitable.  Just
because something is expressed in source format doesn't give it
religious immunity, especially if it needs 18 parts.  It had better be
good, and general.

Unfortunately, none of this gets binaries off the hook.  The problem is
twofold -- technical and political.  Technically you have the difficulty
of assembling working binary versions from the uuenc/shar'd/split-up
format we use; the difficulty of getting binaries to work on clone
hardware and unanticipated system configurations (common problems);
and the overwhelming difficulty getting corrections made when bugs are
spotted.  (You can't do it yourself, the author has to, and authors get
awful busy even when they're being real nice!)

The political problem is more profound, and has to do with a consensus
of what USENET is.  There is a powerful and traditional school of thought
that says USENET is a medium for stuff humans can read and understand:
text articles, tables, poems, source programs and such-like.  Binaries
aren't readable.  There is also a school of thought (the "programmers
rule!" school) that says USENET readers ought to be able to *tinker*
with anything they grab from the net.  You cannot easily tinker with
binaries.

Now it turns out that programmers, for all their vaunted concern with
the source code, are USERS too just like the rest of us, and almost
always use software for which they don't have the source or have never
read the source.  Life is too short (and the employers' personnel dollar
too dear) for everyone to compile their text editor before using it,
for instance.  You just use it.  If someone offers you a small, terrific
utility that increases your productivity regardless of how the source
code looks, you take it.  USENET can be as good at this as the next
net if it wants; the overall traffic shouldn't be too big.

The problem is size.  The hugest stuff just doesn't belong.  And, of
course, some people have 0.00 use for PC anything.  That's why I
recommend (a) a 64k limit on binary postings (couting all "part" articles)
and (b) a separate bin.* hierarchy so that non-PC sites can filter out
the binary stuff in a straightforward way.
-- 
Tom Neff			UUCP: ...!cmcl2!phri!dasys1!tneff
	"None of your toys	CIS: 76556,2536		MCI: TNEFF
	 will function..."	GEnie: TOMNEFF		BIX: are you kidding?