lightdm or gdm?

[Robert Ancell]

Thanks Seb,

= Disclaimer =

Firstly, I should disclose my interests - I am the founder and maintainer
of LightDM. I'll try and minimise any biases I might have. Ultimately I
want the best outcome and I'm happy with either way we go as long as it
makes sense.

= History (simplified) =

In the beginning there was one Display Manager - XDM. Various desktops
forked this project to produce their own display managers that displayed
appropriate UI for their desktop (e.g. Qt on KDE with KDM, GTK+ on GNOME
with GDM).

LightDM was started for two reasons:
- We were having quality / maintenance issues with GDM in Ubuntu. GDM was
highly complex.
- None of the display managers allowed you to change greeter technology
outside of a bit of themeing.

The goal was to have a common display manager daemon (LightDM) and allow
each desktop to develop their own greeter against a stable greeter API
(liblightdm).

Ubuntu switched to LightDM in 11.10 and developed Unity Greeter. I gave up
pushing LightDM in GNOME (which I regret) amidst the Unity-GNOME flamewars.
KDE came super close to picking LightDM as a replacement for KDM with a
huge list of positives (including having the second most active LightDM
developer being from KDE). The major point against it was the CLA. That was
very disappointing. Most of the small desktops seem to have chosen LightDM
and seem quite happy with it.

= Pros and cons =

In addition to Seb's list:

* LightDM
+ an extensive test suite that contains 370 test cases. The test cases run
a full LightdDM setup and simulate all components.
+ a sophisticated configuration system that allows packagers and sysadmins
to override configuration. Some sysadmins may have work in switching from
LightDM to GDM (I can't quantify how significant this is).
+ Extensive logging that helps in debugging issues.
+ Support for greeters in multiple languages / platforms (C, Qt, Python etc
via GIR, Vala).
+ VNC support.
+ Remote login support (this has not been used for sometime and may or may
not be useful in the future).
+ Support for Mir (I don't know if this is important anymore, but dropping
LightDM would mean no DM supports it).

* GDM
- no significant tests (make check runs one thing from 2007)

= My conclusion =

My preference is to continue with LightDM for the following reasons:
- I think the tests and features of LightDM are superior to GDM (I could be
biased here).
- To switch would create some work for sysadmins who use existing features.
- I think the amount of work to get LightDM to work with GNOME Shell is
roughly equivalent to getting new features into GDM.
- LightDM is not an enormous amount of work to maintain - it's quite mature.
- If we were to stop developing LightDM this would be a huge loss to the
community outside of GNOME. Being used across various projects means we
have a larger pool of developers testing and fixing bugs. I'll have to
defer to my manager(s) to decide if this investment / divergence is
considered worth it.
- We always have the option of switching to GDM in the future if we want
to, the switch back would be harder.

Finally a clarification - when you refer to "LightDM" this never means any
UI that you see (LightDM is a daemon and contains no UI code). So
continuing with LightDM does not mean using Unity Greeter. The Unity
Greeter code has bitrotted to such a point we clearly would not use it (it
was to be replaced with Unity8 Greeter).

--Robert

On Thu, Apr 20, 2017 at 2:32 AM Sebastien Bacher <seb128 at ubuntu.com> wrote:

> Hey there,
>
> That's a topic that was mentioned during the meeting yesterday and
> something we need to decide on early in the cycle since there is going
> to be work needed on that front to have a fully working session.
>
> I'm doing a small summary of what I think are the pro (+) and con (-) of
> each
>
> * lightdm
>
> + well tested in Ubuntu
> + we have people in the team knowing the codebase
> + shared with other flavors/greeters selection
> + guest session
> - divergence from upstream
> - we are the maintainers so it's more work for us
> - gnome-shell uses gdm for its lockscreen so work is needed to make it
> work with lightdm
>
> * gdm
>
> + that's the GNOME solution, works today with wayland & gnome-shell
> - we started lightdm because we found the gdm codebase not easy to work
> on, that might still be true
> - ?keeps an active session from the greeter even after logging which
> uses resources? (it was mentioned on IRC, to be confirmed, is that
> needed due to the lockscreen?)
> - no guest session, we need to work on that or decide to drop the
> feature from Ubuntu
>
>
> I talked a bit with Robert yesterday who said he could make lightdm use
> the gdm greeter (he has some work started on that a few cycles ago)
> which means it could be used as GNOME lockscreen instead of gdm. He's
> probably the best placed to comment on the work and pro/con of the
> solutions though so I'm going to let him go into the details when he
> replies.
>
> Cheers,
> Sebastien Bacher
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-desktop/attachments/20170419/bc69caa0/attachment-0001.html>