[prev in list] [next in list] [prev in thread] [next in thread]
List: kde-devel
Subject: Re: CORBA on kde-core-devel
From: Simon Hausmann <shaus () uermel ! Med ! Uni-Magdeburg ! DE>
Date: 1999-09-18 6:57:15
[Download RAW message or body]
On Fri, 17 Sep 1999, Alex Hayward wrote:
> On Fri, 17 Sep 1999, Matthias Elter wrote:
>
> > Am Fre, 17 Sep 1999 schrieb Martin Konold:
> > >
> > > This is my main concern. Imagin me clicking on a URL and because konquerer
> > > thinks it is very intelligent tries to load some alpha quality code in
> > > order to display an image or play some unneeded sound. A shlib approach
> > > would kill the complete konquerer instead os simply producing a garbage
> > > picture or playing unpleasant sound.
> >
> > That may be true, but what do you suggest? A buggy CORBA component running in a
> > seperate process is not much better as you can't try/catch every CORBA call.
> > Server dies -> object is nil -> MICO exception -> main app crashes!
>
> I don't see why you can't... You'd check all return codes and make sure
> that you carefully parsed the output of a program you'd forked off so why
> not be similarly careful when calling external code in a somewhat
> different way?
>
> If you are doing something as simple as rendering an image or playing a
> sound that can just simply be aborted if things don't work out then you
> could at least catch CORBA::SystemException, spit out a diagnostic and
> carry on. Couldn't you?
Now seriously, putting a try/catch combination around every CORBA method
invokation *is* nonsense IMHO (even with the help of macros) .
This is possible in libkded/kded, because *there* you know about the *few*
remote invokations, but it is not acceptable in applications like
Konqueror or whole libraries like KOM/OpenParts.
The reason is so simple I think: We use CORBA because it *hides* the fact
whether an object is local or remote. And we would be trying to make a
distinction again.
There is no such distinction, and that is good. Just have a short look at
Konqueror: CORBA allows us to have *no* distinction whether an embedded
view is local or a remotely embedded one. Or just attempt to count the
endless amount of invokation in KOM or OpenParts themselves...
(or go over to KOffice..)
Did you get the point? :-)
Bye,
Simon
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic