[EMAIL PROTECTED] wrote:
On 5 Jun 99, at 8:23, Paul Derbyshire wrote:
> If you check the source code...
I don't have it.
> And, because of a "feature" dating back to days when exponents were
> small, none of the test data sets in the program self-test would
> trigger the bug. So "normal" testing procedures failed to detect it.
The self tests go to arbitrary levels of precision! I've seen it. Before
beginning an exponent it does a self-test for that precision, if it hasn't
already done one there.
> a) This incident was well discussed when the bug was detected
> about 9 weeks ago.
I wasn't there then. Because of my FUCKING ISP and their worthless mail
system...
> b) I can assure you that George did not feel at all good about it at
> the time. Nevertheless, accidents happen. Always have done,
> always will do. That's life!
Accidents will happen. This is why they invented crash test dummies...
I think I'll be staying away from any ".0" versions in future.
> (Though, as I say, if you saw the offending code section, I
> think you'd need to think hard about it for a few minutes before you
> could see that there was a potential problem).
I'd bet you're right. Because, although I am a programmer myself, I don't know
the code like the back of my hand. I would hope each person that does work on
the code does know it like the back of his/her hand...
> d) There is a maxim amongst conservative computer users, "don't
> fix what ain't broken". There was no _need_ to upgrade to v17,
> unless you wished to run double-checking assignments.
I was running double checking assignments until I got the new K6-2 400MHz CPU.
> ...the setback was actually only four or five weeks.
Sounds like a lot to me. This CPU I have can finish LL tests in that kind of
time, except that I have a habit of using it for other things a lot of the
time.
____________________________________________________________________
Get free e-mail and a permanent address at http://www.netaddress.com/?N=1
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
