Re: LSB summit in Boston

[Brian Cameron <Brian Cameron Sun COM>]

Richard:

> The "Linux Standards Base" is a plan to develop a specification for
> the GNU system.  Not, in this case, for the GNU/Linux combination,
> just for GNU, because these specs don't concern the kernel, Linux.
> It is purely for GNU, but they call GNU "Linux".
>
> If we want to develop specs for the GNU system, let's organize it
> under the auspices of GNU.  I have good relations with some KDE
> developers.  We can set it up.

The deeper issue here is interface stability, and I think that we
need to have a good understanding of where we are, and where we
want to go before we can move forward.

Personally, I think the LSB project is not the ideal avenue for
the short-term.  Their requirements for test cases for each
interface is causing them to review older versions of the GNOME
libraries, not the latest versions that most modern distros ship.
Over time, the LSB may well catch up, but this is likely going to
be a process that takes years to complete.  I welcome the
attention they are giving to helping make the desktop interfaces
more stable, but I think it is really the desktop module maintainers
who define what stability means.

At Sun, we are taking the position that a core set of GNOME Platform
libraries are Stable, including atk, pango, glib, GTK+, and
at-spi (needed for writing acessible technology, or AT,
applications).  Since platform libraries have stability
guarantees and the maintainers of these modules have a good
history of keeping the interfaces truly stable, Sun is
comfortable with this.  In addition Sun believes the
FreeDesktop specificiations for Desktop file integration, and
MIME integration are stable, and we are considering making the Icon
Integration specification as Stable.  We think that this set
provides a reasonable set of interfaces for anyone to
write an application and integrate it into the desktop.

If the GNOME community were comfortable with this sort of stance,
then there is really no reason why the GNOME desktop couldn't claim
to support this degree of integration as Stable today.  In fact, I
would argue that we already do.  As we approach our GNOME 2.16
release, there is no history of serious interface breakage.

The biggest problem with GNU desktop stability is, in my
opinion, an issue of communication.  The GNOME and FreeDesktop
projects do a poor job of communicating to its users what interfaces
should be relied upon.

I think some work is needed in order to truly provide a welcoming
environment.  Most importantly:

1. The FreeDesktop specification page is a hodgepodge of different
   specifications, and it isn't clear which specifications are
   "blessed".  Note that most of them are at pre 1.0 versions.  I

   The FreeDesktop specification page says "freedesktop.org is not a
   standards body. The naming of this page is a historical accident,
   and will take a bit of wiki lifting behind the scenes to fix."

   "http://www.freedesktop.org/wiki/Standards

   Obviously this doesn't inspire a great deal of confidence.  I
   understand that the Portland project is working on this, but
   progress seems to be slow going.

   http://portland.freedesktop.org/wiki/

   I believe there are only a handful of specifications where most
   end users require stability, and there should be a focus on
   making a more formal commitment to these.

2. Note that GNOME System Admin Guide,

   http://www.gnome.org/learn/

   You will see that quite a few interfaces that are exposed are, in
   fact, not specified anywhere.  If the GNOME project thinks it wise
   to recommend users to use these interfaces, they should probably be
   specified somewhere with some sort of assurance that the interfaces
   will not break on upgrade.  Or, if there is concern that they might
   break, this should be clearly communicated as a warning.

   Many of these exposed interfaces are GNOME specific.  More effort
   could be made to determine if these interfaces should be specified
   in a desktop-neutral (FreeDesktop) fashion so the same interfaces
   work for all desktops.

   For example, application configuration integration.  Currently
   GNOME and KDE have different configuration management systems.
   It would be nice if end-users could use a single interface for
   changing and accessing their configuration settings for all
   applications.  In my opinion, this is perhaps one of the biggest
   interface disconnects on the desktop.

3. The GNOME and KDE communities should make more of an effort to
   explain to end-users the advantages and disadvantages of choosing
   one desktop over the other, and provide guidance regarding which
   interfaces should be used by end-users.

   For example, are all GNOME Platform libraries recommended for
   end-use, or are some of them really intended only to be used within
   the GNOME consolidation?  I would argue that really only a subset
   of them should be given a recommendation for use.  This is why Sun
   only suggests end users depend on some of them.

I think the bulk of these issues are related to the overall lack of
good communication between KDE and GNOME, so if you feel that you
would be able to help get the right people together to work on these
sorts of problems, then that would be great.  Perhaps if you wanted
to get involved with the Portland project, that would be a better
forum for making progress on this topic.  I think the FreeDesktop
group may be the closest thing today to a "GNU" standards body that
you are looking for.

Brian