[rfc] The Big Picture

[Samuel Cormier-Iijima]

On 10/26/06, David Zeuthen <david at fubar.dk> wrote:
> On Thu, 2006-10-26 at 21:28 -0400, Samuel Cormier-Iijima wrote:
> > <sarcasm (or is it?)>
> > To DBus people, every problem is a DBus problem.
> > </sarcasm>
>
> Seriously, I think most people are aware that D-Bus is not the right
> solution to every problem on the planet. For example that's why HAL on
> Linux is using a good old Unix domain socket to get events from udev.
> I'm sure you can find other examples.
>
> Also, I think Havoc described the purpose of D-Bus and the message bus
> architecture pretty well... earlier today [1].
>
> I'm not sure what you want to achieve saying what you just said
> however.... did you have a real point or are you just trolling?

I just wanted to make a point... I was thinking about something like
this a few days ago; what if all daemons exposed a D-Bus interface
that could be called (with appropriate permissions) from userspace?
Things such as NetworkManager, gnome-power-manager, HAL all do this
already and seem to be work well. However, I think that e.g. running X
as two separate processes communicating events about mouse motion (!)
through D-Bus wouldn't be the "ideal desktop," as it would slow down
the computer alot. Anyways, even if the desktop was build like this,
what benefit would it give to the user that doesn't already exist?
Perhaps if the original poster mentioned some use cases, I would
consider the idea more. Sorry if it sounded like a troll :-)


>      David
>
> ps. It's D-Bus.... not DBus, D-BUS, dbus or whatever. I'm also looking
> at you Havoc :-)

I've tried to fix it :-)

> [1] : Havoc wrote something like this
>
> "In fact, the _primary_ purpose of dbus for the desktop in my mind isn't
> IPC at all, but all the policy around it - activation, lifecycle
> tracking, session scoping, security policy, and yes the machine id.
> These things provide fixed context and semantics that otherwise don't
> exist. IPC is easy; open a socket and dump stuff into it, there are a
> thousand protocols and libs for this, dbus offers nothing new."
>

Hmm... although D-Bus is so convenient for IPC :-)

Samuel Cormier-Iijima