Re: Bonobo Unique Application v3.1

[Michael Meeks <michael ximian com>]

Hi Gustavo,

On Tue, 2003-04-22 at 15:05, Gustavo J. A. M. Carneiro wrote:
>   Just so I understand your idea, BonoboSignalServer would replace
> BonoboApplication and BonoboSignalClient would replace BonoboAppClient?

	That was the initial plan.

>   I get the impression that with your approach the message sending
> interface would get more complicated.  You would have to register 
> signals for each message type you want to send, and g_signal_new is not
> trivial, plus you need to generate C marshallers, which implies adding
> stuff to the makefile, etc., and people seem to not like this sort of
> thing very much.  g_signal_emit is not elegant either.  Are you sure
> this signal stuff is recommended?  The existing API is more
> programmer-friendly, IMHO.

	Yes; I think you're right.

>   I think people want a more runtime oriented and simple message API.
>   For real typesafe and elegant message passing, we already have
> CORBA. :)

	Heh ;-) I suppose so; I just read it more carefully, and I'm thinking
that the signal emission way is ok. It would be nice if we could have a
set of generic closures that people could register against messages -
but then we have some nasty type safety problems; unless they register
the argument types as well ... one thing to do is to add expansion space
in the class vtable - and in the IDL file [ 8 methods or so in each ].

	I'd also like to see the common case (of passing a set of arguments
from argc, argv to the remote end and determining if the thing succeeded
made much simpler - ie. so we can replace the gnome-terminal code with
something far smaller of the order of a couple of lines.

	I'd also like a hook to be able to add a 'listWindows' thing; so we can
register a method in libbonoboui ( or perhaps we want a separate,
aggregated BonoboUIApplication Interface) that would get a list of all
the UIContainers we had associated with BonoboWindows - so we can start
to do more interesting automation type tasks.

	Anyhow - I think it looks good in the current form; and I'm writing a
GEP to get the ball rolling on including it.

	Regards,

		Michael.

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