Re: [kaffe] Planning for 1.1.2 release - XAWT obsolete?

2003-09-23 Thread Clemens Eisserer
Hi there!

Nope. GNU Classpath AWT integrating should happen for the 1.1.3.

Does this mean the current kaffe-xawt implementation will be thrown away?

This was the best awt implemenation I´ve every seen, and I dont 
understand why gnu-classpath isnt based on this awt.
The only problem are the ugly widget LF´s, but  should be possible to 
make this whole thing skinable without big troubles.

Sad :-(

lg Clemens

___
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe


Re: [kaffe] Planning for 1.1.2 release - XAWT obsolete?

2003-09-23 Thread Dalibor Topic
Hallo Clemens,

Clemens Eisserer wrote:

Nope. GNU Classpath AWT integrating should happen for the 1.1.3.

Does this mean the current kaffe-xawt implementation will be thrown away?
I hope we'll have some way of interfacing to 'kaffe-style' AWTs from the 
past as well. We'll be able to see what can be done, when James and 
Stephane have finished their efforts to port kaffe's Qt-Awt and Xlib-Awt 
over to Classpath's peers. I hope we'll at least end up with some docs 
on migrating your peerless AWT to GNU/Kaffe [1].

The intention for kaffe's AWT is to transform it to work with 
Classpath's peers. It shouldn't be a problem to ship the 'old, but 
working' AWT alongside it, and provide a compile time selection switch 
--enable-old-awt that would compile the AWT libraries from different 
directories.

This was the best awt implemenation I´ve every seen, and I dont 
understand why gnu-classpath isnt based on this awt.
A) Incompabile licensing. Classpath is GPL + linking exception, kaffe is 
GPL, so Classpath can not include code from Kaffe.

B) Copyright assignment. The author of kaffe's AWT, Transvirtual, 
doesn't exist anymore, so there is noone who can transfer their 
copyright to the FSF anyway.

Anyway, if you want to help make the transition to Classpath easier, 
you're more than welcome to start looking at kaffe's and Classpath's 
classes and to point out where things can be taken over from Classpath 
already without breaking kaffe's AWT in any way. I've been doing this 
off and on when I carved enough time for it, but my time is 
unfortunately quite limited at the moment. So I'd very much appreciate 
any effort to enhance what we have in kaffe already by replacing more of 
the old code by code from Classpath where there are no negative effects.

cheers,
dalibor topic
[1] Bad joke, I know.

___
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe


Re: [kaffe] Planning for 1.1.2 release - XAWT obsolete?

2003-09-23 Thread Jim Pick
On Tue, 23 Sep 2003 18:47:28 +0200
Clemens Eisserer [EMAIL PROTECTED] wrote:

 Hi there!
 
 Nope. GNU Classpath AWT integrating should happen for the 1.1.3.
 
 Does this mean the current kaffe-xawt implementation will be thrown away?

Not at all.  We already support multiple AWTs -- this would just be
another one, I imagine.

 This was the best awt implemenation I´ve every seen, and I dont 
 understand why gnu-classpath isnt based on this awt.

I think it comes down to ownership/licensing dynamics (eg. Transvirtual
never transferred copyrights to the FSF for the AWT).

 The only problem are the ugly widget LF´s, but  should be possible to 
 make this whole thing skinable without big troubles.

Yeah, the LF is definitely ugly.  :-)

I haven't done too much AWT stuff myself, but I think we could serve as
a useful place to experiment with AWT integration since we have somewhat
looser licensing constraints than other projects.  Here's what I'd like
to see:

 1) Multiple AWT implementations

 a) Some using lightweight widgets, where the widgets are drawn in
Java.  eg:
 
- our current X AWT
- Swing (we can run Sun's old AWT based one, and should be able
 to run their newer Java2D based one if we get a Java2D
 implementation)
- others, eg. Rudolph (from Wonka)

The Transvirtual version of Kaffe had a lot of additional AWT
work done on it.  The big thing was it had a lightweight AWT that
could render to a no-native-wm backend, which provided a pluggable
backend for the core graphics primitives on a number of platforms
(X, Linux framebuffer (using the fgl library), Wind River's UGL,
GRX, MGL, Allegro).  The no-native-wm backend sounds similar in
concept in many ways to what Java2D does - it might make sense
to move that stuff into a Java2D implementation.  There is probably
additional window-manager functionality that Java2D doesn't cover.

Transvirtual's AWT and Swing both provide support for skins / LF's.

As far as backends go, I think it would be cool to experiment with
even more extreme Java-only solutions, eg. driving a framebuffer as
a Java byte array, or writing a xlib completely in Java!

Graphics backends are really interesting, because the challenges
involved in utilizing modern graphics acceleration hardware are
interesting.  Things like Cairo and DirectFB on Linux, DirectX on
Windows, and Quartz on Mac OS X provide use ways to get access to
accelerated hardware.

If we had a really good Java2D implementation, it might even be
feasible to use something like WeirdX as a more flexible
replacement for XFree86 to support legacy apps.  :-)

b) Some using native widgets.  These provide tighter integration
   on some platforms.

- Gtk - nice for Gnome desktops.  There's a version of Gtk which will
   render to the framebuffer too.
- Qt and Qt/Embedded
- Win32 (for Windows)
- there was a MicroWindows AWT for Kaffe based on the Win32 widgets,
  I think.
- a Carbon based AWT for Mac OS X would be nice
- Motif/Lesstif would be interesting, just because it's what Sun used
  in their original AWT
- SWT apparently works, which works with Win32, Motif, and Gtk.  It
  would be cool to do a SWT backend that used lightweight widgets.

   Note that some of these widget sets are themeable themselves, and
   can use multiple backends as well.  eg. Qt is essentially a lightweight
   widget set that renders it's own widgets on the platforms it supports.
   Conceivably, it could even be made to use a Java2D backend.  :-)

  2) Compile-time switching of AWTs

 All these AWTs should be configurable at configure time.  For
 embedded platforms, it's nice to be able to select only one specific
 AWT tuned to the target platform.  Transvirtual took this to extremes
 with fgl, where even the color depth was chosen at configure time.

  3) Run-time switching of AWTs

 It would be nice to be able to configure Kaffe with multiple AWTs,
 and have them selectable at runtime using a property.

 If somebody wanted to do something even more extreme, I imagine that
 it would be possible to create and API to enable applications to use
 multiple AWT implementations simultaneously.  :-)

Sascha Brawer is organizing a free Java AWT BOF at Linux Kongress - that
should be interesting:

  http://www.linux-kongress.org/2003/bofs/index.html

Anyways, I don't think there is a one-size-fits-all solution for
graphics on Java - I'd like to keep an open mind, and hope some really
nice interfaces and implementations evolve.

Cheers,

 - Jim

___
kaffe mailing list
[EMAIL PROTECTED]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe