Re: GNUstep and the desktop

[Alexander Malmberg]

Fred Kiefer wrote:
> Looks like I did misunderstand your proposal. I was thinking about well
> defined interfaces.

The ultimate well defined interface here is the public interface for
these classes/methods. The interface between -gui and desktop bundles
should be as close as possible to this interface to provide as much
flexibility as possible for desktop bundles.

> But from this latest mail it looks like you don't
> want teh functionality at all.

I want the functionality. :) However, I, and others, want it in many
different environments, and the right way of providing the functionality
varies wildly between environments. Thus, I don't want the
implementation in -gui.

> Could you please explain again how GUI
> would interact with the external components than?

Which external components? For the surrounding environment, and stuff
like gsnd and a workspace application, that answer is that -gui wouldn't
directly interact with any of them. -gui would, at startup, load a
"desktop bundle". This bundle handles all interaction with these
external components.

In your "pure GNUstep, OPENSTEP-style" desktop bundle, you can implement
it by communicating with gsnd and a workspace app. Other desktop bundles
may use other strategies.

> To me this is either
> via a well known interface or not at all.

The interface between -gui and desktop bundles would be well defined.

> Lets take the workspace or the sound as an example. What is wrong with
> the way we have split up the functionality right now? (Apart from the
> horrible looking code that starts the sound server)

In NSSound? The current split assumes that sound is handled by a daemon
(not true for windows), and that this daemon can be communicated with
using DO (not true for GNOME).

- Alexander Malmberg