[Zope3-dev] Re: proposed changes to contained helper functions

[Philipp von Weitershausen]

Fred L. Drake, Jr. wrote:
> Shane Hathaway writes:
>  > Giving a unique name to each child serves the purpose of making all 
>  > children potentially addressable via URL.  That's one of the sweet spots 
>  > of Zope; it makes the system discoverable and clear.  I think a decision 
>  > to not use this pattern should be weighed carefully.  In particular, if 
>  > an object is not addressable via URL, it is quite difficult to provide a 
>  > Web user interface for it or any of its descendants.
> 
> He's right not to use names here; the DOM is well-defined.  Exposing
> individual node of the DOM directly via Zope is questionable (IMHO),
> but I don't know if that's what he's doing.
> 
> I think he just wants to use part of the published interfaces in his
> internal structures.  I hope that's all this is about.  ;-)

If this is about ILocation/IContained, then the answer is no. I would 
like to use some general machinery which, of course, would certainly use 
  interfaces. I could imagine the following ones, for example::

   class IChild(Interface):
       __parent__ = Attribute(...)

   class ILocation(IChild):
       __name__ = Attribute(...)

   class IContained(ILocation):
       pass

Adapters that are about parenthood could be registered for IChild rather 
than ILocation, for example. That's the "machinery" I'm talking about :).

Philipp