>=20
>=20
> 7/11/03 8:07:36 AM, Alex Kilpatrick <AlexK@HCITRAINING.COM> wrote:
>=20
> >I do care about it.  But since the second board works, I=20
> know it is not=20
> >a problem with the design.  My design is not crashing, a particular=20
> >implementation of the design is crashing.
>=20
> That doesn't follow.  It's entirely possible that something=20
> about your design causes it to work under very narrow,=20
> "ideal" conditions but fail when certain hidden assumptions=20
> aren't met.  A correct design works *all the time* under the=20
> conditions that *you explicitly specify*; it doesn't rely on=20
> any parameters that are just taken for granted.
>=20

OK, maybe "I know it is not a problem with my design" is too strong of a
statement.  "I have more evidence that it is not a problem with my
design" is more appropriate.

But you say a correct design works "all the time" under conditions that
I explicitly specify.  For the purposes of the demos that I am going to
do, those conditions are:

1)  Connected to my laptop
2)  Using my cables
3)  In the conference room
4)  With a brand new battery
5)  With me running the demo, and no you can't touch it until I am done
with the demo

Obviously, these are pretty ideal specifications.  In fact, I found out
that a regular serial cable is much more noisy than a USB to serial
cable during a recent demo. =20

Testing takes time.  The more general your specifications, the more
testing you have to do.  You can trade off those two things to get the
level of reliability you want.  It isn't a binary thing.  If I told my
client that I had tested the device enough to have confidence that it
will work in a 100 degree temperature range, with ten different kinds of
power supplies, in five different levels of EMI, they would slap me on
the wrist for wasting time with that stuff.

I will certainly concede that there are good design things that take
essentially zero resources (like adding decoupling caps), and I do all
of those things that I know about.

Alex

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics