[SLL] Arguments against unknown process of open-source: Where is Eric Raymond?

[Robert Woodcock]

On Tue, Apr 15, 2008 at 06:49:06PM -0700, Xeno Campanoli wrote:
> I now allege that the problem is we open-source / free-software folks 
> don't think about the specific substance and falsehood of this argument. 

Well, if the goal is reliable, trustable software, then I see several
factors, each of which on their own can only *prevent* the software from
being high-quality:

1. The software's author(s) must thoroughly understand the task the software
   will perform and the way that users expect the software to help them.
2. The software's author(s) must thoroughly understand the languages and
   tools they are using to make the software, and the environment the
   software must work within.
3. The software's author(s) must take personal pride in what they are
   building.
4. Each of the software's features must be used by a lot of people (the
   more, the better).
5. Those users must have the ability to recognize a bug in the software for
   what it is.
6. Those users must be able to report a bug to the software's author(s), in
   a way that either allows the author(s) to see what happened, or to
   replicate the bug on their own.
7. The software's author(s) must have the time available to maintain the
   software after its release.

Note that only #1, #2, and #3 are measures of the author(s) skill.

#1 is probably the most common reason for a code rewrite. Obviously #1, #3,
#5, and #6 are best served if the author and user are the same person. #4
affects a lot of very expensive vertical market software. #6 is an Achilles
heel of most commercial software, and economic factors often doom #7,
especially for one-off custom code. #5 often makes #4 a law of diminishing
returns.

Usually, what you have is a festering mess which is deficient with respect
to several of these factors. Usually, at least a few are beyond your control,
and what's worse, attempts to improve on one often have consequences for
others. In other words, you can put in all of the business processes and
metrics you like, but "unknown process" will, in the end, carry the day -
regardless of your coding methodology.
-- 
Robert Woodcock - rcw at blarg.net
"When Mrs. Pattycake comes to us to be taught, turn that wandering doubt in
her eye into a fixed, dedicated glare and she'll win and we'll all win.
Humor her and we all die a little. The proper instruction attitude is,
'You're here so you're a Scientologist. Now we're going to make you into an
expert auditor no matter what happens. We'd rather have you dead than
incapable.'"
	-- L. Ron Hubbard, Scientology Operating Thetan documents