Re: X and other visions

[Michael Banck]

On Sun, Jun 13, 2004 at 12:05:50PM -0700, Thomas Bushnell, BSG wrote:
> Michael Banck <mbanck@gmx.net> writes:
> 
> > Note that Desktop Environments such as GNOME are not Linux specific.
> > GNOME runs (and is well supported by Sun) on Solaris, for example. Of
> > course, GNU/Hurd might have a better implementation of virtual file
> > systems as Linux, but nevertheless you need a platform-independant
> > fallback if you're running on something else than GNU.
> 
> FWIW, this I think is one of the chief problems in operating systems
> design today.
> 
> Any popular and full-featured interface is going to need to run on
> vanilla (plain posix, X, etc) and cannot rely on whatever fancy tricks
> the kernel-level services can provide.  And all the various user-land
> vfs things for desktops demonstrate that it can be done well without
> any kernel-level fancy tricks.

Well, having a well-designed, portable user-space solution is surely
nice. But replacing parts or all of it with superiour kernel-level
implementations (if available) would make sense as well, just like glibc
has general routines and hardware-specific optimizations.

Perhaps the GNU project should have thought about a portable
virtual-file system library in the spririt of gnulib back then when it
become at least possible that the Hurd was not ready as basis for the
desktop. Perhaps GNOME is too far detached from the GNU project for any
serious code-sharing/interoperability (cf. e.g. the lack of copyright
assignments and virtually no awareness of the Hurd from the GNOME
people).

> Truly we do have a problem now that emacs and gnome and kde and web
> browsers and so forth all have their own *different* ideas about how
> to do vfs-stuff and how it should look, but that's a problem that can
> be solved by coordination, 

As far as the desktop (I know Emacs can be used as a desktop, but my mum
uses GNOME and I won't get her switched to GNU Emacs anytime soon) is
concerned, integrating layers and common things are developed by
freedesktop.org these days and not the GNU project (e.g. HAL and D-Bus
for hardware abstraction, the menu system, the gnome-system-tools
backend was proposed to be moved there recently). It seems the GNU
project has more or less lost the leading position in innovation and
guidance, if it ever had it. The people at fd.o are not bad and quite
clueful WRT Free Software, but I'm afraid that in a couple of years the
only thing most people using a Free operating system will see of the GNU
project are a couple of lines at bootup. 

Just like it is totally hilarious SPI had GNOME as one of their projects
up at their website, it's unrealistic that GNOME is still considerd to
be a GNU project these days, apart from the name.

> and cannot be solved by fancy kernel-level tricks--since all those
> userland tools will need to work, and work consistently, on platforms
> that don't have the cool tricks.

The GNOME project seems to prefer usability over portability with the
assumption that the other platforms will catchup eventually. Thus,
gnome-volume-manager (based on the project utopia hardware abstraction
stack) will probably be included in GNOME-2.8 although it is Linux-2.6
specific. It is unfortunate that GNU/Hurd was not able to jump in when
the need for something like libgnomevfs came up and has to catch up now.

Those who do not understand the Hurd have to reimplement it poorly :)


Michael