Re: New module proposal: LightDM

[Brian Cameron <brian cameron oracle com>]

Ray:

On 10/22/10 09:50 AM, Ray Strode wrote:
> > However, not everyone really needs or wants the degree of integration
> > that GDM provides with GNOME.
> I feel like GNOME should be catering to GNOME's users, and we're doing
> a disservice to them if we don't provide integration.

Agreed, and I agree that GDM should focus on providing this integration.
Overall, it does a great job at that.

> > It should, I think, be possible to define a lighter degree of GNOME
> > integration for a display manager that intends to be a bit more desktop
> > neutral/agnostic than GDM, but still be a part of the "GNOME Desktop".
> I disagree.  I think that's going the wrong direction.

I fully agree with you that having a tightly integrated display manager
like GDM makes a lot of sense.  I would go so far as to say it makes a
lot of sense for the most common use-cases (e.g. single-user
desktop/laptop environments probably being the most common).

However, to suggest that GDM is the best solution to all problems and
use-cases is probably overselling.  Some examples:

- non OpenGL

  GNOME is moving towards OpenGL with GNOME 3.0.  The GNOME community
  recommends using gnome-session/gnome-panel for the "classic" look.
  I imagine GDM will eventually go the clutter/OpenGL route.  Does GDM
  plan to always support non-OpenGL environments, or is there value in
  having a lighter display manager to focus on this sort of use case?

- non-Linux

  GDM and ConsoleKit are very Linux-focused and include a fair amount of
  code that is hard to get working on non-Linux distributions.  It may
  not be reasonable for GDM to position itself as the best display
  manager for all operating systems.

- Multiple Desktop Environments

  In environments where a distro allows the user to configure what
  desktop to use on the system, GDM may not make sense since it
  requires so much of the GNOME stack to be installed.  If the user
  decides to not install GNOME, then using a display manager that
  does not require so much of the GNOME stack to be installed is a plus.

  Perhaps the GNOME community doesn't really care so much about this use
  case since we might not care what non-GNOME users use.  But, I am just
  highlighting that some distros may choose to not use GDM (or not use
  it all the time) just because they might want a display manager that
  is not quite so integrated with one particular desktop.

- Security/Paranoia

  In general, the more code that runs at login time creates more
  opportunity for malicious users find ways to compromise the code.  I
  think that the GDM team has done a great job locking things down so
  that it is reasonably secure.  However, it is impractical to audit
  everything GDM now depends upon (metacity, gnome-session, gnome-
  settings-daemon, GConf, etc.) so it is hard to know for sure.  The
  fact that this infrastructure is constantly churning (e.g. metacity
  to mutter, GNOME 2 to GNOME Shell) means that new issues could pop-up
  anytime on upgrade.

  This degree of paranoia is probably not important to most GNOME users
  who just want a login screen to keep random people from logging into
  their laptop.  However, there are environments where security is more
  of a concern and where users would want to trade some usability for
  more security.  It is easier for a lighter display manager to be
  audited and to provide more reasonable assurance that the code is
  stable and secure simply because considerably less code is running
  at login time.

Just a few examples of use-cases that may not be in-line with the
direction of the GDM project.

> > Think about Ephipany.  Users who really want tight GNOME integration
> > may prefer Epiphany, but some users may prefer other web browsers for
> > various reasons.
> I don't think that's a fair card to play.

I was not playing any card.  I was just trying to provide an example
of another situation where GNOME supports multiple programs with
differing degrees of GNOME integration.

> > I thought the new release model was all about choice and flexibility.
> Nope, I think you misunderstood (or I did).  The new release model is
> about putting even apps on an even footing.  It doesn't apply to the
> core of the OS.

While I agree that the display manager is more "core" than the average
application, I do think there are plenty of reasons that could cause
a user or distro to decide what display manager makes sense in different
environments.

> Users should be able to pick which apps they want,
> not which window manager, settings daemon, or login window they want.
>  From a desktop point of view, these things are central to defining
> what the GNOME is.  They are the "OS" which defines the stuff around
> the apps.  Our mantra should be "integration, coherency, consistency,
> just works" for the OS.  Adam Jackson did an awesome email a while
> back called "Linux is not about choice" that's mildly relevant, so
> I'll post it:
>
> https://www.redhat.com/archives/fedora-devel-list/2008-January/msg00861.html

Not everybody who uses GNOME uses Linux or shares this mantra.

> We really need to band together and focus on making our OS better, not
> more flexible, just better.  I don't want to ship a bucket of parts to
> our users and leave them to fend for themselves. I want the user to
> get something refined, cohesive, and out-of-their way so it fades into
> the background while they're trying to get their job done (whatever
> job that may be).   They shouldn't have think about the computer when
> they use the computer as a tool to do something else.   We're not
> there yet.  It should be an important goal.

Of course.  I do not anticipate that most users would care to change
their display manager.  Most users would likely be happy with their
distro's choice.  So this is probably not an issue.

Instead I would think that only distros or particular projects with
display manager integration needs might reconfigure the user's system
to use the display manager that works best in these environments.

> The only way we'll get
> there is if we work less on modularizing and more on integrating.

The "only" way?

> > > Anyway, I'm obviously in favor of keeping GDM in GNOME.
> >
> > So am I, but I'd think there should be room for more than one display
> > manager choice in GNOME.
> Again, I disagree here for the reasons mentioned above.  We need to
> focus, not splinter.

If the GDM community really thinks that supporting such non-typical but
important use-cases in GDM is important, then it would be great to find
better ways to make GDM more of a community-focused project where it is
easier to innovate changes that falls outside of the most common
uses-cases.

I do think that there are some real use-cases where a lighter display
manager choice adds real value.  Especially if the GDM community is
not so interested in "polluting" GDM with code to support uncommon
but important use-cases.

Brian