From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : GGI mailing list <ggi-develop@eskimo.com>
  Date: Wed, 5 Aug 1998 16:40:08 -0700 (PDT)

Message for GGI folks... (fwd)

	Josh asked me to forward this to the list.  Although I think he is
panicking a bit too much about our "window of opportunity", he has some
good points.  Things are getting too chaotic, our documentation is
slipping further and further out of date, and things are way too messy.  I
think we really need to tighten things up here.  I'll try to update the
FAQ sometime soon. 

Jon

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

---------- Forwarded message ----------
Date: Wed, 5 Aug 1998 17:57:08 -0500 (CDT)
From: Joshua M. Yelon <jyelon@cs.uiuc.edu>
To: taylorj@ecs.csus.edu
Subject: Message for GGI folks...


Could you please forward this to the ggi list?

-----

I notice that the GGI project is busily missing a window of
opportunity...  the kernel folks finally have graphics support in the
form of fbdev, and everyone wants to play with it.  Lots of people who
weren't interested before are now looking for demo programs that use
graphics.  The GGI project could easily use this opportunity to grab
some good publicity.  But the GGI project has made it so difficult to
get libggi that it's practically impossible to actually try libggi.
Here's the ordeal I had to go through to finally run a demo program:

Step one, I downloaded the advertised release of GGI, version 0.0.9.
I didn't get as far as compiling it, a quick glance told me it was too
old.

Then, I went to the snapshots directory and downloaded a snapshot as
of yesterday.  I looked inside and found the fbdev stuff.  I tried to
compile it, but the fbdev target wouldn't compile.  I played with the
source code, and couldn't figure out what was wrong... it looked like
it was only half-finished.  Eventually, I figured out that I had an
out-of-date copy.  As it turns out, the snapshots aren't being
updated... it may be ``yesterday's snapshot,'' but it was a snapshot
taken yesterday of a tree which hasn't been updated in a month.

I realized I'd have to download and install CVS to get the real thing.
I downloaded CVS and read a little of the CVS documentation.

I then managed to check out a copy of degas.  I went into the libggi
directory, configured and compiled as instructed.  It failed during
``make install.''  After some experimentation, I figured out that
``make config'' hadn't set one of the configuration parameters
correctly.  I manually edited ``.config,'' and recompiled.  This
produced a library which installed correctly in /usr/local/lib.

I then compiled the demos (easily), but when I tried to run them, they
complained that they couldn't find libggi.  I figured that this had
something to do with ``ld.so.conf.''  I added /usr/local/lib to
ld.so.conf, but that didn't help.  So I started over, and recompiled
the whole thing to install in /usr instead of /usr/lib.  This produced
working demo programs (from inside X).

However, the programs were installing their own colormaps, grabbing
all 256 colors, and making the rest of the screen unreadable.  I
figured that was probably configurable, and I knew (from observing the
mailing list) that libggi has environment variables which control it.
I looked around for documentation about the libggi environment
variables, but never did find it.

Finally, I tested it using a frame buffer device.  It complained
about ``unable to open default visual.''  I figured that that meant
it was using the wrong display target.  I wasn't sure what to do
about that, but after grepping the source code for a while,
I found a reference to the "FRAMEBUFFER" environment variable.
I took a wild guess and did a "setenv FRAMEBUFFER /dev/fb0".  That
did the trick.

Bear in mind, I'm somebody who already knows a lot about libggi.
I've been reading the mailing list for months, and I have a great deal
of background knowledge.  If I had been a true novice, my ordeal would
probably have been twice as difficult.  That was way too hard!!!

If you want people to use this thing, you have to make it available!
It pains me to wonder how many hundreds or thousands of people have
downloaded libggi and given up on it.  How many opportunities do you
want to miss?  Remember, you're not alone in the world... if libggi is
impossible to install, people will write for svgalib, or for OpenGL,
or for X windows.  I figure you have about two more weeks during which
the novelty of fbcon gives you a good window of opportunity.  If you
don't make an impression, it's back to obscurity for the lot of you.

Here's what you need to do: write a new web-page.  It should contain:

  1. A ``.deb'' and a ``.rpm'' containing libggi compiled for
     X and fbdev targets.  Post both libc5 and glibc versions.

  2. A separate ``.deb'' and ``.rpm'' containing the demo programs.

  3. Instructions for setting environment variables before running
     the demos, and how to run the demos from inside X.

  4. Instructions for how to compile a kernel with vesafb or vgafb support.

  5. Precompiled modules for the kgicon drivers that people can
     insert into their kernels.  Make sure you support several different
     kernel versions.

Yeah, it's a pain in the ass... but do you want people to use this
stuff, or NOT!?

Hide the old web-page.  Especially the obsolete documentation, and the
silly diatribes about why KGI is so great.  You don't want people to
bore people, you just want to tell them how to install it, and give
them the tools they need to do it.

- Josh