From:  der Mouse  <mouse@Rodents.Montreal.QC.CA>
	Date:  Mon, 17 Mar 1997 11:42:39 -0500 (EST)

> If I were handed this job, I would proceed as follows:

> - Figure out how to memory-map the framebuffer.

That code is easy to lift out of the current server.

> - Figure out how to turn keystrokes and mouse events into something
>    comprehensible.  This probably means stealing chunks of code from
>    elsewhere, such as the kernel keyboard handler.

The input side of things is relatively easy to lift out of the current server
or out of HPUX server sources, which is where I extracted it from in the first
place.

> - Take a Sun server (sparc or sun3, not much difference).  (Why Sun?
>    Simply because that's the one I know best, so it would be the
>    easiest one for me to figure out what each piece of the code does.)
>    Clone it into a new ddx tree.  Rip out all the hardware interface
>    stuff and shove in code from the first two steps.

The difficulty here is the *byte* per pixel addressing of the topcats and
catseyes (even mono topcats) which just doesn't map onto the model that the X
server uses hence the modified mfb and cfb directories that the server uses ---
and the need then to rework those changes into new mfb and cfb code each time a
new X release comes out.  This is what the HPUX R3 server did (last one HP
released source for) which is what this server was based on in the dim distant
past.  With R4 they managed to hide the addressing differences within the HP
code and use the standard cfb and mfb but they never released the source (which
is why I wrote the server in the first place) so I don't know how they did it.

cheers
mark