PolicyKit integration in KDE take 1: systemsettings and kcmshell

[Dario Freddi]

Hello guys,

As you might know me and Daniel got polkit-qt and polkit-kde ready for the 
masses. This is only the first step though, since now the real integration can 
begin. And before we start pushing out tutorials (just give us time to find an 
easy and bullet-proof system and you'll get them, we are still experimenting), 
I wanted to talk you about what seems the most obvious implementation realm: 
KCMShell and SystemSettings (*shake*)

Let's face it: the old administration button was a ugly and unsecure hack. We 
don't need to start a GUI as root, instead we will delegate an helper to do 
that. The model will be the following:

 - Developer x develops his nice KCM module that needs saving as root, just as 
if this wasn't needed.
 - In the save() function, instead of saving, he starts a DBus-activated 
helper that will verify if the caller is authorized to perform that action 
through PolicyKit (that will manage eventually authentication on its own). 
That part will be covered by the tutorial.
 - Everyone wins: developers just need to split their code to a separate small 
app (and we are planning to provide interfaces and eventually cmake macros for 
it), no more hacks, everything is more secure, and users will have their 
functionality back.

Now, everything seems so cool, but there is one small details. I am already 
planning the patch to both systemsettings and kcmshell, but I have some doubts 
about implementation that I wanted to share with you. Let's go:

 - How to implement this on the GUI side? I thought about two options: 1) just 
make the "Apply" button call PolicyKit if needed 2) Add an "Unlock" button 
that makes all the widgets in the module active (just as GNOME does). I 
personally favor for the first, if we can get some nice icons (think about the 
shield in Vista).
 - New KCModule interface. I plan to add a function setPolicyKitAction(action) 
that, when called, takes care of modifying the GUI as above and makes 
authentication for action needed for saving. I think this is the easiest 
approach, but I appreciate opinions.
 - Saving. We could make save() get called only if authorization is obtained, 
or add a saveOnAuthorization() method that gets called only if authorization 
is obtained, while save() will get called in any case. I strongly favor for 
the first as it would make things a lot easier for everyone.

I'd like to hear thoughts, suggestions, ideas, whatever.

Thanks and cheers everyone
Dario

-- 
-------------------

Dario Freddi
KDE Developer
GPG Key Signature: 511A9A3B


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20090309/65279065/attachment.sig>