29 Aug
2010
29 Aug
'10
7:51 p.m.
I posted this last Wednesday - but for some reason, it didn't make the
"cctech" list. I'm posting it again FYI...
Cheers,
Lyle
---------- Forwarded Message ----------
Subject: Re: Should you correct production mistakes?
Date: Wednesday 25 August 2010, 17:20:22
From: Lyle Bickley <lbickley at bickleywest.com>
To: "General Discussion: On-Topic Posts Only" <cctech at classiccmp.org>
On Wednesday 25 August 2010, Eric Smith wrote:
I wrote:
To be specific, during the original restoration, we checked every
solder joint in the PDP-1. We found four bad connections. We then
reviewed the implications of the bad connections in terms of function
and reliability. The Team decided to only repair two of the connections
because they could affect normal operation and demoing of the system.
We left the other connections "loose". As Eric already stated,
we documented everything and labeled all changes (fixes) we made.
BTW: We have never experienced a problem with the "unfixed" solder
joints ;-)
Cheers,
Lyle
--
Lyle Bickley, AF6WS
Bickley Consulting West Inc.
http://bickleywest.com
"Black holes are where God is dividing by zero"
Had we found any problem analogous to that in
the PDP-1 restoration,
where the machine operated correctly despite a manufacturing defect, I'm
sure we would have had a debate on whether to fix it. I think my own
opinion in that case would be that we should leave it alone, but tag the
module (paper tag attached with a short loop of string) and document the
issue in the system logbook.
Fred Cisin wrote:
In the example that I mentioned of an IBM MDA
board, it worked for about
10 years, and then became intermittent. Even that, I would NOT consider
to be "operates correctly", as the particular defect IS a defect, and may
eventually bring the system down.
I think the way that I would state it is that just because the system is
working correctly (at the moment) doesn't mean that it isn't broken.
Part of the reason I don't think we'd fix such a problem in the PDP-1 is
that the PDP-1 isn't doing anything criticial. We can afford to have
downtime if a latent problem eventual causes a failure. If we've
properly documented that latent problem, we can check for it when the
system does fail, and fix it if necessary at that time.