Re: Bag IDL

[Michael Meeks <michael ximian com>]

Hi Mike,

On Sun, 25 Feb 2001, Mike Kestner wrote:
> Miguel de Icaza wrote:
> > This was specifically done this way to *show* that PropertyBags were
> > supposed to have EventSources, and that it is not a second-hand hack.
> 
> In other (OO) words, it's a kluge for:

	I'm getting quite annoyed with this. I have answered you multiple
times.

	a) It describes the contract
	b) We do not have a service description concept
	c) We have frozen bonobo so we cannot impl. b)
	d) It makes things easy for language bindings:

		a.getEventSource.addListener () instead of
		a.queryInterface("IDL:Bonobo/EventSource:1.0").addListener()

	e) I dislike the lazy 'X' interface naming prefix.

	Ergo it is not a kludge.

> I still contend that it would be more appropriate to document the
> *Events* that an Implementation of the PropertyBag Interface should
> emit.

	No one is arguing with this whatsoever. They need documenting,
preferably automaticaly. This cannot be done in the IDL currently - and
again, Bonobo is frozen.

> Putting an extra QI function in the vtable does little to help the
> component developer or user understand the functionality provided by a
> component implementing PropertyBag.

	It specifies a contract that will always be applicable to this
interface, there is no problem with this being in the interface, and QI
can cause pain for bindings cf. the MS Chart interfaces eg.

	Regards,

		Michael.

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