[Chicago] python 3.0 hep
mtobis at gmail.com
Mon Dec 8 17:36:16 CET 2008
I think Dijkstra's opposition to testing is wrong in favor of proof is
In practice, often all you can do is say either you got the same
answer Donna did, or not. It is true that you cannot prove you will
always get the same answers as Donna would without a complete formal
model of what Donna does, and that sometimes neither Donna nor her
organization are capable of producing such a model, so Dijkstra would
presumably shrug, call matching the exiting test cases the desired
predicate, and end up with what we call tests.
As you say, "try it and see" is good advice, along with "failing
that, try finding someone with more experience on this sort of thing
and put up with them sneering at your ignorance", both forms of
courage which the professional programmer would do well to summon.
Still, I would think that Dijkstra's attitude has value. For the most
part if you have to resort to guesswork or calling on someone who
knows some secret workarounds or investigatory tricks, some other
software professional has failed to provide you with a reasonable
platform for your own work.
Install hassles are especially huge sloppiness multipliers. I have
finally developed some of the skills to cope with installation hassles
but I resent every minute spent on "learning" those "skills" or coping
with the problems they address.
In the case of Python, I think every install hassle I've seen relates
to multiple Python installations on one platform. I can generally
muddle through nowadays but I don't think that such experience is
valuable to my skill set as a developer. Newbies should not have to
deal with such an experience, and that includes newbies building from
More information about the Chicago