[LightDM] Hard dependency on AccountsService

[Yves-Alexis Perez]

On lun., 2011-08-22 at 11:26 +1200, Robert Ancell wrote:
> On 20 August 2011 02:53, David Edmundson <david at davidedmundson.co.uk> wrote:
> > Do you have links to anything explaining about AccountsService?
> > Explaining what it is, why it's better than dealing with files directly,
> > that sort of thing.
> > Dave
> 
> I talked to Ray Strode who is one of the main developers of
> AccountsService.  Some relevant quotes:
> "AcccountsService has always been an ad hoc "serve our needs right
> now" type thing.  We don't have much documentation about it."
> "Ideally at some point it would become integrated with or superceded
> by SSSD at some point in the future." [1]
> "It's main purpose at the moment is to serve the needs of GNOME, but
> the kind of thing it's doing is general, so it made sense to keep it
> on freedesktop so it could get adapted as needed for other desktops if
> there is interest."
> 
> (I think this is an appropriate position to take given the age of the project)
> 
> Unfortunately there isn't much I can point you to.  I'll add a Wiki
> page on freedesktop.org that at least provides a basic overview so
> there is something to point at.  From my point of view the advantages
> are:
> - Puts all behaviour in one place - which means less code in lightdm
> and less places for bugs
> - Caching/storing of information done in one place - currently the
> .dmrc may be unreadable by different users, there needs to be a cache,
> home directories may not be mounted...
> - Consistent API - all user accounts features are accessed using
> D-Bus, so don't have to use glib for some features, open files, run
> programs for others
> - Support for new backends - who knows what sorts of account systems
> we will migrate to in the future (e.g. Facebook logins?), a good API
> means we don't care

Oh dear...

> - Able to modify users using the API - it may make sense to add users
> from the greeter, so a service like this would make this much easier

I' not really sure that's something I want.

> - Easier to test - you can make a mock accounts service easier than
> faking all the different components of the current system
> 
> I think based on this and the feedback received lightdm will
> definitely continue to support both legacy and accounts service for
> the near future, but I'd like to hope we can move to an API like this
> in the future.  So please get this idea on the agendas of the
> different desktops :)

Please, KISS. I'm not really sure I like that AccountService stuff if it
means overengineering things.

Besides, right now I'm not sure it's really relevant for a cross desktop
and cross-distro tool. 

The switch to systems libraries to avoid code duplication is nice, but
right now it seems that keeping dmrc files as a fallback looks like a
good idea. (and note that there are other areas where code duplication
could be dropped, like for Xauthority generation).

Regards,
-- 
Yves-Alexis

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/lightdm/attachments/20110826/caa25510/attachment-0001.pgp>