Call for testing: LightDM

[Robert Ancell]

On 06/07/2011 08:03 PM, Matthew East wrote:
> On 7 June 2011 10:02, Alan Bell <alanbell at ubuntu.com> wrote:
>> yeah, I would very much hope that lightdm does not introduce more
>> accessibility regressions.
> I'm taking this opportunity to post a link to this comment on the
> proposed switch to lightDM from Matthew Garrett, in case people
> reading here haven't seen it, it seems relevant to this discussion and
> I haven't seen it mentioned before.
>
> It also briefly discusses impact on accessibility, albeit without
> going into detail.
>
> http://mjg59.livejournal.com/136274.html
>
Needless to say, I disagree with most of the points Matthew has raised.  :)

The argument that LightDM does less because it is smaller is weak.  The
features listed as missing are the ones currently provided by GNOME
session.  However, you could very easily write a LightDM greeter that
ran a GNOME session (i.e. just copy the relevant code from GDM) that
would provide all that same functionality for almost no significant
additional lines of code (as the lines are all external modules).  So
LightDM could provide all the same functionality as GDM with it's
current size.  However, if you produce a greeter that does implement
this functionality by the numbers given in the blog post you would have
a whopping 49,000 lines of code to use before you became bigger than GDM.

My preferred design is not to use GNOME session, with the main reasons
being startup cost, complexity and the security risks of running a full
session in a login screen (as pointed out by Chris earlier in this
thread).  However, the assumption seems to be this will involve
rewriting every service.  Not the case; of course a GNOME greeter would
leverage as much as is appropriate of the GNOME platform e.g. using
gnome-power-manager if that was the best solution.

A point that is incorrect (which I pointed out to Matthew but he didn't
correct) is things like power management are not performed in the
backend.  So the backend is not going to swell to support different
desktops.  Sharing policy between login screen and session is up to the
greeter.  A GNOME greeter should have the same policy as a GNOME
session.  It wont be the same as KDE policy.

Finally I think Matthew massively underestimates the value of being able
to differentiate on the greeter.  A number of projects - including
Ubuntu - have wanted to have a greeter that matches their desktops but
have been unable to do so with GDM.  No-one really cares (or wants to
know) how the lower layers of display management work (I certainly
didn't), they just want to work on the GUI.  If we were to modify GDM to
provide the UI we want (has been attempted a few times) we would be
carrying a huge patch on top of the already 35 we are carrying in
Natty.  Using LightDM we are able to run the daemon unpatched and
differentiate to our hearts content with the greeter.  With GDM we would
effectively be forking the project.