[lug] can't make this stuff up, folks...

Davide Del Vento davide.del.vento at gmail.com
Tue Oct 27 15:57:39 MDT 2009

Excellent points. Let me share a story, just happened.

If somebody drives a big commercial truck on a pedestrian bridge, and
the bridge collapse killing several people, it's not considered a
failure of the design.

While I was doing some tests on our hw, I needed to plot "some" data.
I ended up writing up my python scripts using one of the
best-respected plotting libraries in python.
Note: in this case, I was the customer of myself, telling the other
me: "I need this plot, just make it work!"
My scripts worked fine. Until "some" data became more than 2GB (the
amount of physical RAM), when the script became much slower (it
started swapping like crazy) - not really user friendly, but I didn't
tell "the customer", since he just wanted an image. My scripts worked
ok. Until eventually I didn't get any plot at all when my "some" data
reached 4GB (I was plotting on my 32bit laptop). To get the plots in
time to the f*** customer, the less painful solution was to install an
Ubuntu 64 bit on a spare desktop (with 1GB of  physical RAM), enabling
the largest available partition as swap (48GB) and let it chew for one
night-per-plot. It was shaky, but it worked, like the pedestrian
bridge on which somebody drove a large, commercial truck and luckily
it doesn't collapse

The SE in me wants to blame the customer, but he really couldn't
figure out at the beginning that the data would have grown so much (I
trust him, being myself :-) And even if he could, that wouldn't help
much, since this particular plot is an histogram, you are actually
counting "how many things" are there and you aren't supposed to keep
'em all in memory: you just read the file, and count 'em.
But unfortunately the fine library requires that everything is in memory.

I thought this is a great little example that shows why software is
different. A pedestrian bridge is not a highway bridge (in costs,
size, and kind of passers). A histogram plotting library should be the
same (and actually is not more difficult) for pedestrians or trucks.
Other kinds of plots might not be the same - fair enough - and thus
the library's author(s) just used the same approach for the histogram


> Sorry for this late reply, I've been out in the sage brush sea (Nevada) and
> had no connections.
> If an automobile breaks down on a bridge causing a two to four hour traffic
> jam, it is not considered a failure of the design.
> If an untrained driver, drives off the side of the bridge, it is not
> considered a failure of the design.
> If some one purposely damages the bridge (vulnerability) it is not
> considered a failure of the design.
> Just my thoughts,
> Steve

More information about the LUG mailing list