Archived from groups: alt.games.everquest (
More info?)
On 19 Dec 2004 15:53:58 GMT, johndoe@example.com wrotC
RIVE_E
>What else could happen that would require a team of hardware engineers
>40 hours to figure out the problem? I'm a software guy, buy I've seen
>hardware guys at work, and usually they can identify issues pretty
>quickly using self-diagnostics and such (assuming you're using "real"
>hardware and have a "real" support contract).
I think what you need to also consider is the "Try
this..test...nope..." cycle. For some systems, running through the
checklist of fixes takes a few seconds to a few minutes each. Other
times, it takes hours.
A program I worked on would sometimes manifest bugs only under extreme
-- but real world -- conditions. It was a financial simulator, and
clients would set up runs for a weekend and then come back on Monday.
If what they came back to was an error screen, I heard about it. Big
time. In-house testing usually used ten minute to 1 hour runs; for
obvious reasons, testing EVERY feature in a three day run was not
practical. We'd set off one "big run" just to be sure, but since there
were thousands of options, we couldn't test every combinatin of
settings to be 100% sure one particular blend -- often with only one
type of data -- wasn't going to crash it.
If it takes even a half hour to apply a patch, boot, test, and try
again, even a smallish checklist of "stuff to try" can take a long
time. THEN, you have to apply it on every server, and check them all.
This isn't "Windows crashed, reboot, what's the big deal?" stuff here.
This is uber-complicated. (I know YOU probably know this, but a lot of
people with no experience in complex networked systems think their
methods of dealing with a single-system glitch scale effortlessly to a
server farm running an astoundingly complex progrma.)
And bugs aren't always evident. A long time ago, I was a 4D programmer
with Peat Marwick. One of our divisions constantly lsot data with the
program I was working on. I eventually tracked it down to the fact
they had entered "Aetna insurance" with the "AE" ligature character,
instead of "A" "E". 4Ds indexing for text couldn't handle high-ASCII
in a field. Kaboom. This never showed in any test cases, because we
never used high ASCII in names.
People who call themselves "programmers" because they took a Visual
Basic class during summer school very rarely have any grasp of what
coding is like in the real world, and have ridiculous expecations of
both the predictability of bugs and the difficulty of fixing them in a
timely manner.
*----------------------------------------------------*
Evolution doesn't take prisoners:Lizard
"I've heard of this thing men call 'empathy', but I've never
once been afflicted with it, thanks the Gods." Bruno The Bandit
http://www.mrlizard.com