Newsgroups: news.software.b
Path: utzoo!henry
From: henry@zoo.toronto.edu (Henry Spencer)
Subject: Re: (C News) limit filesize and inews
Message-ID: <1991Jan6.073729.21354@zoo.toronto.edu>
Organization: U of Toronto Zoology
References: <1991Jan4.202944.7348@zoo.toronto.edu> <3461@gazette.bcm.tmc.edu> <1991Jan6.004313.8792@zoo.toronto.edu> <3463@gazette.bcm.tmc.edu>
Date: Sun, 6 Jan 1991 07:37:29 GMT

In article <3463@gazette.bcm.tmc.edu> sob@tmc.edu (Stan Barber) writes:
>>>To take the ivory tower attitude that the end-user should get the OS fixed
>>>it commendable, but not always reasonable.
>>It is both reasonable and necessary in certain cases, unless you have
>>unlimited time to spend and don't care about the impact on future maintenance.
>I think that's what I said. However, if you post code that should be widely
>usable and it doesn't run on the most popular varients of Unix available, then
>I have to question if that is due to the inability to get on a system to test
>the code or the "ivory tower attitude" that the varient is broken and the 
>providers of the varient should fix their code in order that your code should
>work.

Popularity has nothing to do with correctness, as various companies have a
habit of demonstrating.  I'm sorry, but when a certain popular release of
Unix miscompiles cleanly-written code if and only if cc's -O option is
used, to the point where the code fails spectacularly, then I don't *care*
whether I had such a version available for testing -- this is a *BUG* and
I refuse to contort my code to allow for it.  Same for when a certain large
manufacturer releases a shell in which

	x=y
	if test " $x" != " y"
	then
		echo oops
	fi

prints "oops":  I'm sorry, this is just plain wrong, and I refuse to
spend hours mangling dozens of shell files to avoid provoking it.  I
will document the problem -- and a look at C News's notebook/problems
will reveal that these are real examples, by the way -- but that's all.

It is not a question of "they should fix their code so that mine will
work":  they should fix their code so that *their customers'* code will
work, because it is most unlikely that these bugs affect only my code.
I don't care how popular these bugs are, they remain inarguably and
unambiguously bugs.

This is not an "ivory tower attitude".  Quite the contrary; I consider
it an ivory-tower attitude -- thoroughly impractical and not reflecting
the real world -- to say that "you should make your code work on every
popular system, no matter how badly broken; the only excuse for failure 
is not having had a system available for testing".  In the real world,
there always comes a point where you have to say "Enough!  Munging my code
to allow for these morons' mistakes is not worth the time and trouble.".
The costs are real and must be taken into account.

(Stan, I don't understand why with one breath you agree with this, and
then with the next breath you say "if it doesn't work, then either you
didn't have one to test, or you have an ivory-tower attitude".  Whether
I had one to test is utterly irrelevant to whether the changes are too
costly to be worthwhile.)
-- 
If the Space Shuttle was the answer,   | Henry Spencer at U of Toronto Zoology
what was the question?                 |  henry@zoo.toronto.edu   utzoo!henry
