From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : GGI mailing list <ggi-develop@eskimo.com>
  Date: Mon, 6 Jul 1998 00:57:48 -0700 (PDT)

Last kgicon update before it goes in CVS

	Like the title says, this is it.  I am not going to accept any
more patches or release any more tarballs myself.  This will give us
incentive to get kgicon into the CVS tree quickly so everyone can work on
it without having to go through me.  I am quite sick of staring at this
code and am going to take a breather for at least a few days.  Well, let's
see if I still feel this way when I get up tomorrow |->.  

	I wanted tonight's release to be v1.0, but as most of the drivers
are not yet converted and there are some bugs left, I can't legitimately
call it that.  Oh well, one more week+weekend should allow us to get
kgicon into the CVS tree, fix the remaining bugs, convert the remaining
drivers and work with 2.1.109, which should have a slew of fbcon bugfixes
and new features that were just released in patch form by Geert and Martin
Mares.  That should result in a v1.0 we can announce on linux-kernel and
be proud of.  Hopefully kgicon will go into the kernel in .110 or .111.

	Anyway, here's what's new in this release:

* Converted ET6000 driver (Andrew)
* Converted Matrox drivers (Jonas)
* Fixed physical vs. virtual framebufer address (Andrew)
* Accel directory remakes properly (Andrew)
* Set correct bitfields for 16, 24 and 32bpp in set_color_bitfields() (me)
* Cleaned up old code in kgifb_init() (me)
* Removed some old CVS/ directories that I missed before (me)

	I broke 32bpp somewhere |-<.  I haven't been able to track down 
where this is going wrong yet.  That's the only outstanding "bug" that I 
know of, but there are still many missing features (|->):

* The rest of the drivers need to be converted.
* Any remaining cruft from the old KGI that isn't being used should be 
  removed in the drivers and makefiles.
* Any remaining naming inconsistencies (i.e. ggi -> kgi) should be 
  cleaned up in the drivers and makefiles.
* The use of .inc files should officially be deprecated (that is, their 
  use should be discouraged but conversion should not be mandatory) now that 
  the Matrox drivers have shown that they are not needed anymore.
* Only the default mode can be set.  The other multimode fbcon drivers 
  build mode tables - we could simply do this too, but it would not
  take advantage of KGI's flexibility.  We should figure out how to do this
  on the fly somehow.
* No ioctls are handled.  Few fbcon drivers handle ioctls either, but 
  that is no excuse for us |->.  Also we need logic in the ioctl handler
  to recognize KGI ioctls and call the appropriate kgim_*_command() 
  functions. 
* The default mode needs to be requested from the KGI driver instead of 
  fixing one default mode with fixed timings.  
* True textmodes should probably be used if they are available.
* Panning/ywrap should be used if the driver supports them.
* Make multiheading work properly.
* Code up assembly-optimized sotware text rendering routines.
* Code up accelerated text rendering routines.
* Code up heuristics to make optimally-balanced use of the above.
* Make KGI drivers compile into the kernel correctly.
* Make sure that drivers work properly on non-x86 architectures.
* Add KGI ioctl support to the fbdev LibGGI target.
* Backport the existing fbcon drivers to KGI to show people the inherent 
  superiority of the KGI way of doing things.
* Implement Marcus' unified monitor driver and remove the monitor subsystem.
* Implement monitor config via /proc.
* Implement /proc/kgi.
* Many other great ideas I can't think of right now.

	The future is very bright for KGI!  Let's get out there and show 
the world what we can really do!

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed