Re: Shipping Vera with 2.4
- From: Havoc Pennington <hp redhat com>
- To: Michael Meeks <michael ximian com>
- Cc: GNOME Desktop Hackers <desktop-devel-list gnome org>
- Subject: Re: Shipping Vera with 2.4
- Date: Thu, 27 Feb 2003 12:00:35 -0500
On Thu, Feb 27, 2003 at 10:58:26AM +0000, Michael Meeks wrote:
> As for making GNOME into an entire operating system - there's a straw
> man if ever one saw one. Simply looking for sensible areas of
> re-applicability for various technologies is great: glib has a use
> outside GNOME - so does gtk+, so does ORBit2.
I'm all for that. I think the real point is simply that things should
be in the right place in the dependency layer cake, considering the
whole big picture of an operating system with an X11-based desktop.
That means assuming a world larger than GNOME (including VCL, WINE, Qt
based apps) and it means keeping things properly modular, and properly
> How does eg. having standardized GUI tools across distributions with a
> sane structure make GNOME an "operating system" ?
What I mean by "making GNOME an operating system" is the mentality
that GNOME is going to drop in and have features that should be on the
OS level but instead apply only to GNOME apps. For example,
GNOME-specific font system is just crack. A GNOME tool that
*configures* systemwide non-GNOME-specific font configuration is a
good idea. See the difference?
We *need* Qt and WINE apps to integrate pretty well. Yeah, a GTK app
might have a more consistent UI in many details and that's hard to
fix; but there's no excuse for having a different list of fonts
available in these different apps, or for having to set up my text
editor in 6 places.
For example, CUPS and fontconfig are getting good traction across
all desktops and distributions/OS's, and so users have the same fonts
and printers in all their apps. That's an example of the right thing.
The wrong thing would be GNOME-lpd, or GNOME-fontconfig.
> One of the things that amazes me about D/BUS is that - having been
> patronizingly lectured extensively in the past about how "a string is an
> API" - it seems that D/BUS is essentially a "send a string" transport.
> Is there really no formal contract specification language ? and/or
> suggestions for and/or descriptions of such ? I couldn't find such a
> thing in the documentation. You're going to need IDL - hey, you could
> even compile that IDL to some typesafe stubs / skels !
There probably will be an IDL, yes, and typesafe stubs/skels, on the
level of the GLib/Qt integration libs.
] [Thread Prev