Fwd: User bus conclusion

[David Faure]

On Wednesday 10 November 2010, Kevin Krammer wrote:
> On Wednesday, 2010-11-10, Chani wrote:
> > > > Actually, with the Akonadi based kmail, it no longer needs to be a
> > > > singleton and in fact isn't anymore, afaik. There is a per user
> > > > Akonadi server, and you can connect to it from as many clients,
> > > > graphical or otherwise, as you wish.
> > > 
> > > In theory.
> > > When you have a second D-Bus sessiom bus, running an Akonadi
> > > application will attempt to start Akonadi for that bus, which will
> > > fail because the database is already being used the Akonadi on the
> > > first session bus.
> > > 
> > > A user bus would allow Akonadi to be shared between user sessions.
> > 
> > I don't know much about dbus, but... wouldn't the application have to
> > know which bus to look on, then? would that be troublesome?
> 
> They currently assume session bus, i.e. while we store the bus address used
> by the server in a runtime file, clients basically just use
> QDBusConnection::sessionBus().
> 
> Assuming the existance of a user bus, this would be the one server and
> clients would connect to, thus avoiding the problem of using the wrong
> session bus.

Yes for daemons a user bus sounds good. But not for anything with a GUI - e.g. 
you wouldn't be able to run any kuniqueapplication twice (once on each 
display), since the bus would tell it "you're already running".

I like the idea of a user bus, but I'm very much against the idea of 
deprecating the session bus. We should be able to use both.

-- 
David Faure, faure at kde.org, http://www.davidfaure.fr
Sponsored by Nokia to work on KDE, incl. Konqueror (http://www.konqueror.org).