Object path values - which service do they belong to?

[Havoc Pennington]

On Wed, 2005-05-18 at 18:09 +0100, Ross Burton wrote:
> However, the returned object implements the Book interface
> so I have to reset the interface:
> 
>         dbus_g_proxy_set_interface (proxy, "Book");
> 

This is the kind of thing I'm talking about. If we really want to make
it look like we're passing objects around, OBJECT_PATH isn't right. 

We need a location-independent encoding of the object (including address
and bus name), and we need to shove "enough" type information around
with the object (such as the interface). There are also issues like
"when are two objects the same object" and object lifecycle and so on
and so forth that start coming up.

Making OBJECT_PATH look like an object is an incredibly leaky
abstraction, I guess I would say. The argument for OBJECT_PATH rather
than CORBA in many ways is that this particular abstraction is
_inherently_ leaky so the best goal is to keep things simple and
transparent.

But if we do make things look like objects, shouldn't we back up and
make the dbus protocol really support plugging the leaks that can be
plugged?

Havoc