Python Virtual File System: A spike test

[holger krekel]

Mike C. Fletcher wrote:
> Hi Holger,
> 
> I really like the idea of the filter package (I'm guessing it would work 
> without alteration even with the spike-test class).

No, not really. Or maybe it would? em, i don't really know but i don't
think so :-)

> As for providing methods with precisely the same names as the os module; 
> I'm not particularly thrilled with the names used there, based as they 
> are on UNIX commands, rather than descriptions of what the method does; 
> with that said, I'm not pathologically averse to including such method 
> names.

I find the stdlib-names not very convincing, either. But having to learn 
new names is a PITA for users of a filename-module. Leveraging
the knowledge of 'os.path' and 'os' names makes a lot of sense to me.
  
> Creating an abstraction for the various URL designations is definitely 
> attractive.  Being able to pass around typed URL objects, rather than 
> raw strings, with rich methods on each type should make the Python 
> programming experience feel a little more consistent.

I agree! But 'simple' should win anytime about 'featureful' but inconsistent.
Finding a unifyied URL-interaction interface seems quite difficult. Up until now i
only worked towards a scheme for unified URLs which didn't provide direct
interaction with the (local/remote) object they point to. This *is*
desirable, though.

> Is your module available somewhere to take a peek?

Hope to have it downloadable this evening (GMT). (need to re-setup my
webserver and understand more of HTML4.1 first :-) I'll send the link
to this thread as soon as possible. 

    holger