[Chicago] python 3.0 hep

Martin Maney maney at two14.net
Tue Dec 9 18:52:59 CET 2008


On Mon, Dec 08, 2008 at 10:36:16AM -0600, Michael Tobis wrote:
> I think Dijkstra's opposition to testing is wrong in favor of proof is
> impractical.

It may be impractical, but he's not wrong: testing is literally trying
to "prove" a negative (the code has no errors) by showing that this
case works, and that case works, ....  Code so simple that exhaustive
testing is actually possible probably has a trivial formal proof - in
both cases it's the usefully complex code that offers difficulties. 

  (Before you say "but I test boundaries, so I don't need to test every
   possible case", consider on what basis you know that that works (both
   in general and, perhaps more importantly, in the specific case). 
   It's because you have some less-or-more formal proof (or, I suppose,
   sometimes just a faith-based belief) that it works that way.  here
   again, it's the interesting cases where you both need the sort of
   certainty you get from a proof and are least likely to have it
   ("proof is hard").)

I agree, of course, that when you don't know what exactly the code is
supposed to do you cannot prove it correct - but that doesn't make
testing the cases you [think yo]u do know about much better.  But it may
be all you can do, to be sure.

> In the case of Python, I think every install hassle I've seen relates
> to multiple Python installations on one platform. I can generally

Speaking, sort of, of installing multiple versions of the snake's
tongue, here's a brief report on the successful effort to port a
sizable project to share a common code base for operation under both
versions 2 and 3:

  http://mail.mems-exchange.org/durusmail/qp/441/

-- 
Threaten not the comic with your lawyers' bluster,
all toothless to suppress parody and satire;
for you will not amuse him, but you may inspire him.
  (me, inspired by http://www.netfunny.com/rhf/jokes/01/Apr/mcrhf.html)



More information about the Chicago mailing list