On Sun, 2008-05-18 at 12:40 +0100, Daniel Drake wrote:
> Bastien Nocera wrote:
> > Heya,
> > 
> > Here's a big patch to add user tracking to the device object. Only the
> > user/process that called ClaimDevice can subsequently use that device.
> 
> I didn't know that dbus exposed user information like it does. Is it 
> secure?
> If so, it seems fine. I previously thought we'd have to use ConsoleKit 
> to determine the user. Any pros/cons of each approach?

It is secure, we can then use PolicyKit to tighten access to certain
calls.

> > I'm also thinking about adding PolicyKit to the whole mix. Should I make
> > PolicyKit a hard dependency, or a compile-time one?
> 
> I think it's fine to be a hard dependency.

Cool.

> > So I think that we should:
> > - kill Claim/Release, and claim/release devices when we actually need
> > access to the hardware
> > - add a single call to change the UID to work on, we would (by default,
> > and probably more finely-grained with PolicyKit support) only allow root
> > to change the active UID, and single users to set it to theirs.
> > 
> > This would allow us to fix the second use cases. With PolicyKit support
> > we could fix the user management tool to not have to run as root, as
> > well as the first use case.
> 
> Sounds fine. Only thing that springs to mind is what happens if you set 
> one user, load some print data, then set another user.

That would only happen if a particular user was allowed to set a UID as
the current one that wasn't his.

>  But I can't see 
> any reason for breakage. I think such an interface would actually allow 
> you to load enrollment data for one user, and then verify it against 
> another user's print. Not sure if that is useful to anyone, but perhaps 
> we can label it as a feature ;)

I'll get started on that then.

Thanks!

_______________________________________________
fprint mailing list
[email protected]
http://lists.reactivated.net/mailman/listinfo/fprint

Reply via email to