Destructor

[Peter]

In article <mailman.1022010522.10951.python-list at python.org>, 
shalehperry at attbi.com says...

> when in doubt fire up the python shell and test (-:

The problem with an approach like this is that it really only 
demonstrates what one particular implementation does. In cases where the 
language definition allows for variations among implementations, it 
would be easy to conclude more than you should about what is guaranteed 
to happen everywhere.

This is not not say that trying things out is a waste of time. However, 
it shouldn't really be considered the definitive statement of language 
behavior. In other words: you can't write portable programs based on 
what a particular implementation does or does not do.

I realize that these comments are, perhaps, irrelevant for a language 
like Python since the reference implementation, to some extent, defines 
the language itself. However, this is a very important issue for any 
language defined by a formal document. For example, the C++ standard 
leaves certain things "unspecified" so that implementations can choose 
different behaviors as appropriate for each platform/market/etc. It also 
allows for language extensions.  Thus if any particular implementation 
exhibits behavior X there is no guarantee that any other implementation 
will do the same. One must consult the standard to know what is and is 
not portable behavior.

Peter