"DBus Embedded" - a clean break

[Philip Van Hoof]

On Wed, 2011-01-26 at 10:29 -0500, David Zeuthen wrote:

Hi,

> On Wed, Jan 26, 2011 at 10:11 AM, Philip Van Hoof <spam at pvanhoof.be> wrote:
> > At Tracker we saw a performance drop of a few percentages after we moved
> > from libdbus to GDBus (and yes, we already use FD passing). You're not
> > convincing me than doing worse than libdbus is a good design goal.
> 
> Dude, I never said I wasn't interested in improving GDBus performance.

Improving GDBus's performance was the only thing you were replying to me
about. You rambled a little bit about 3D engines and mid-90 shader
vintage. But dude, I ignored all that. Whatever with your 3D engines..

I already asked you to relax, dude.

> In fact, I said that I'd welcome this. And I've also said many times
> that GDBus currently isn't optimized at all. For the record, that
> discussion is here

You didn't reply this to me. I'm not planning to spend my entire
lifetime reading all of GNOME's bugs just to make sure that I know all
of what you said once in the past.

>  https://bugzilla.gnome.org/show_bug.cgi?id=634471

Sure.

> But so far the only complaint I've had was a synthetic not-so-useful
> benchmark comparing GDBus to libdbus... that suggests that GDBus is 4
> times slower. And given my view on when and when not to use D-Bus for
> IPC, then we're doing just fine here.

I have nothing to do with this synthetic benchmark. We benchmarked
Tracker ourselves. I wrote "a few percentages", not 400%.

> If you have performance data about GDBus, please do share it in bugzilla.

A few percentages worse than what Tracker had before (dbus-glib with a
few optimizations using dbus-1)

> >> This notion that we absolutely need to fit square pegs in round holes
> >> need to stop.
> >
> > I don't understand this analogy.
> 
> The point is that using D-Bus for all IPC is wrong.

FD-passing, sure.


Cheers, dude


Philip


-- 


Philip Van Hoof
freelance software developer
Codeminded BVBA - http://codeminded.be