[Zope3-dev] Re: ISite misplaced in zope.app.component.interfaces

[Philipp von Weitershausen]

Michael Howitz wrote:
> zope.app.component.interfaces.ISite is used in (at least) zope.location, 
> so the site concept seems to be necessary outside zope.app.

It seem so. Please also see my reply to Roman's email.

> I think, ISite should be migrated to zope.component or even (if it has 
> nothing to do with the component architecture) to a separate package 
> zope.site (which not yet exists).

When we moved stuff out of zope.app, we made the mistake of overloading 
zope.component. I wouldn't want to make that mistake again. That's why I 
don't think it should go into zope.component. I was once close to moving 
ISite out of zope.app.component (and Baiju actually did it once without 
discussing it first, IIRC), but Jim had doubts... I don't remember, 
perhaps he can speak up :)


By the way, I personally find the word "site" a bit misguided. An ISite 
is not a "website". An "ISite" may often be used as the root object for 
a website, but it can just as well be used for other objects in the 
hierarchy.

In fact, technically speaking, a "site" is just a place that has access 
to a component registry. So "sites" are places in your object hierarchy 
that allow component registrations, in other words, that allow the 
alteration of component acquisition. (If you compare that with Zope 2 
and the old-school attribute acquisition, every object was a site back 
then. Now it's limited to very specific objects that provide ISite).

I would much rather call this "place that has a component registry". 
That's a bit too long, of course. "component place", or even better, 
"component site" sounds short enough to me for a package name (e.g. 
zope.componentsite).


To cut a long story short, I'm +1, but zope.componentsite or so would be 
much preferred to zope.site.


-- 
http://worldcookery.com -- Professional Zope documentation and training