Lease vs. Refs ...

[Michael Meeks <michael ximian com>]

Hi Darin,

	I've been thinking about your mail for sometime - despite the
apparent lack of feedback.

	It seems to me that while the 'ref' carries no concept of
ownership - there is often a very clear ownership hierarchy in the
code. For a simple example, consider the throbber - clearly the
process should die when all it's controls have either been freed, or
become disconnected / defunct in some way.

	So - I think, that for many cases we can provide some explicit
and clumsy way of terminating factories when they are no longer
needed. Clearly this is sub-optimal in many ways, but here is one
proposal that we can possibly get done inside the API freeze:

	* Make the BonoboControl imortal when it is associated with a
ControlFrame - the lifecycle ref/unref's on its aggregate no longer
have any effect,

	* Couple it's lifecycle purely to either a) the X plug /
socket or b) the ControlFrame's CORBA reference. In fact we could
probably use both of these I think, one for in-proc one for out.  With
ORBit2 it should be fairly possible to listen for a signal on the
connection - I'll add this, to catch broken connections without
needing to poll.

	Does that sound like a feasible, semi-permanant solution that
has some merit ? 

	Regards,

		Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot