[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: A new framework for embedding ... without CORBA
From:       Matthias Elter <me () main-echo ! net>
Date:       1999-09-29 7:48:18
[Download RAW message or body]

On Mit, 29 Sep 1999, David Faure wrote:
> 
> Our master Torben has once more improved KDE's framework by ....
> rewriting what used to be openparts, without using CORBA anymore,
> just plain shared libraries. The benefits are huge, as you all guess,
> and this is something we should consider very seriously.
> Please note that this DOESN'T mean removing CORBA all together. 
> We still need it for all the other cases where we use it right now :
> talking to a running application (kfmclient<->konqueror/kdesktop, kmail, ...)
> and scripting an application (which is the same kind of thing in fact).
> 
> Also please keep this a bit confidential for now, we all want to have
> a header on slashdot like "KDE drops CORBA", which is not true.
> We simply want to have a better framework for embedding, and to get rid
> of the ugly duplication of kdeui/Qt widgets API in openparts 
> (which created something like another toolkit, half-incompatible with 
> kdeui/kdecore/Qt). With the new framework, this half-baked API
> disappears and everything uses the normal KDE and Qt API.

I had a look at canossa earlier this week and I was impressed by the
blasting fast kspread and the compile times. Great work! But I do not vote for
Torbens proposal for several reasons:

- A canossa OpenParts will be totally KDE specific, I still favour a open IDL
interface. Yes I know it does not look like we may find a way to agree on a
common object model with GNOME in the near future but with a C++ interface
OpenParts we will _never_ see GNOME Workshop (or whatever they call their
office suite) apps embeded in KOffice apps or vice versa.

- The current compile times of canossa are fast but as soon as we add CORBA
interfaces for scripting the compile times are close to where they are now IMO.
So I don't see a big improvement there.

- Yes canossa is blasting fast as no marshalling is necessary and as it's using
shared memory. But I'm not sure if the speed improvements are worth the
cost of dropping the IDL interface. KOffice apps are already quite fast right
now and there are some ways to significantly improve performance without
dropping the IDL interface. Tinymico, cuteidl and a possible shared libs
approach using CORBA interfaces to name what I suggested recently.

Greetings, Matthias

-- 
Matthias Elter
me@kde.org / me@main-echo.net
KDE developer

[prev in list] [next in list] [prev in thread] [next in thread]