🚀 go-pugleaf

In article <bhayden.724690634@teal> bhayden@teal.csn.org (Bruce Hayden) writes
(replying to me):

> Well, the problem is that it is very hard to write software that doesn't
> fail (ever), especially at commercially reasonable prices. So - everyone
> is worried that they will get sued for the failure. This may be because
> software is inhearantly more complex than other forms of engineering
> (at least in my opinion). Ask yourself - how many parts does the averege
> appliance, etc. have? Not that many does it. Now - how about software?
>
> Another problem is that the courts have yet to address the question
> of the normal level of care required in a software program. Is one 
> bug in a 1,000 lines of code inidicative of due care? How about 10
> or 100? How significant is the error? I'm here making it sound
> easier than it really is. More than just software metrics is required.
>
> I am now an expert witness in a software case. My opinion is that the
> software is defective, and that due care was not used in its preparation.
> I think that this case is obvious. Many others are borderline. As a
> plaintiff's attorney, I would argue for a high standard of care. As
> a defense attorney, I would argue for a much lower standard.

I agree that legal liability for buggy software is a knotty problem, but
that's not what's important.  Law suits never improve quality.

It is fair to say that no computer company puts quality first, ahead of
featurality.  Until they do, quality will remain abysmal.

The issue is not whether they exercised "due care" or whether they found
the last bug.  The issue is whether bug fixes for all serious bugs are
provided as a matter of course, and by "serious" here I mean anything that
affects the functionality of the software, and whether product is simply
not shipped with known serious bugs.

Every time I say something like this on the net.  Industry people tell me
that economic realities (according to conventional wisdom) dictate that
deadlines come before quality.  He who ships first gets the customers.

And that's why computers don't work.

(end rant)

-- 
Charles Geyer
School of Statistics
University of Minnesota
charlie@umnstat.stat.umn.edu