General question on D-Bus design considerations

[Lennart Poettering]

On Sat, 21.08.10 10:22, Thiago Macieira (thiago at kde.org) wrote:

> On Saturday 21. August 2010 05.29.58 Havoc Pennington wrote:
> > Only minor things really. GVariant has some type system improvements
> > such as maybe types that could be nice. I would change more about the
> > libdbus implementation, which has various flaws. My opinion is that
> > the protocol itself works very well for what it's intended to be used
> > for. I'm sure there are minor things here or there to improve but you
> > know, it works, which is why a lot of stuff uses it, and it even has a
> > lot of nice aspects. If I do say so myself. Most of the things that
> > could be improved are (in theory) improvable via extensions, too.
> > 
> > The implementation side has been rockier, with the reference libdbus
> > being too low-level / least-common-denominator to make anyone happy,
> > though at the same time those properties led to fairly wide adoption.
> > And the higher-level bindings (for GLib anyway) never got done until
> > recently with gdbus, while the rather questionable dbus-glib just sat
> > there tricking people into trying to use it. So people tried to use
> > dbus-glib (ugh) or libdbus (too low level) and got annoyed with those.
> 
> I'm with Havoc. The only thing I'd change in the base protocol is to make it 
> more friendly to future extensions. Right now, the only reaction to an unknown 
> extension is to disconnect, because it's an unrecoverable error.

Uh? That's not really true. There is proper nego implemented for the
unix fd stuff. Not sure what you are missing?

Lennart

-- 
Lennart Poettering - Red Hat, Inc.