:: Re: [DNG] why is polkit needed?
Top Page
Delete this message
Reply to this message
Author: tekHedd
Date:  
To: Harald Arnesen via Dng
Subject: Re: [DNG] why is polkit needed?
On Fri, Mar 6, 2020, at 12:51 PM, Hendrik Boom wrote:
> On Thu, Mar 05, 2020 at 02:09:37PM +0100, Didier Kryn wrote:
> > Le 03/03/2020 à 23:37, tekHedd a écrit :
> > >
> > > So, I would consider rewriting polkit and dbus from scratch.
> > >
> > > Also, who has time to rewrite polkit and dbus from scratch?
>
> What are the actual requirements for a dbus-like system? Requirements
> that would allow a completely different design?


Exactly. Are there even requirements supporting the current design? Were there ever requirements at all? We can easily see what it does, but it's really hard to determine what it *needs* to do.

Bad sign: You know you've chosen poorly the moment you are simultaneously offering a) broadcast messaging and b) guaranteed delivery.

A google search for d-bus requirements turns up, well, documentation of its current architecture. No requirements. Also contains this choice quote:

"The usage of D-Bus is steadily expanding beyond the initial scope of desktop environments to cover an increasing amount of system services. For instance, NetworkManager network daemon, BlueZ bluetooth stack and Pulseaudio sound server use D-Bus to provide part or all of its services. systemd uses the D-Bus wire protocol for communication between systemctl and systemd, and is also promoting traditional system daemons to D-Bus services, such as logind.[25] Another heavy user of D-Bus is Polkit, whose policy authority daemon is implemented as a service connected to the system bus.[26]"

So... all of the usual suspects. What is absent here? That's right, no *other* programs are listed besides the usual suspects. So who really uses it?

Nothing I can find suggests that dbus is used for anything essential, besides possibly polkit. And there's nothing suggesting that polkit needs to be implemented via dbus. Therefore, you could eliminate dbus entirely and rethink polkit's implementation without undue impact, assuming you are ditching systemd and friends of course.

(I realize I'm skirting "devil's advocate" territory here...)

t