> From: Paul Kranenburg <pk@cs.few.eur.nl>
> Date: Wed, 30 Apr 1997 22:33:21 +0200 (MET DST)

> > Also, a question: is the sparc code using sys/dev/ic/zs8530* ? Or
> > isn't that the same chip? What is the difference?
> 
> Yes, it's the same chip and no, the sparc port does not use that MI
> code yet.
> 
> While I like the improved visibility of configuring devices (keyboard,
> mouse, tty, etc.) on top of a base driver for the chip, I'm definitely
> less charmed of the proliferation of 8530 specific code into the
> top-level drivers, especially the mouse/keyboard front-ends.

Well, I guess I'm responsible for that "proliferation" of 8530 stuff
into the keyboard and mouse drivers, but in my defense I'll just say
that my only goal in that project was to try to divide the existing
code into separate modules (for tty, keyboard, mouse) and connect them
together using the autoconfiguration system.  They are a little better
than the old sparc zs code (lots of stuff in zs.c) but I have never
made any claim of perfection.

(I believe in incremental improvement, etc... 8^)

> Not only does this prohibit attaching a mouse or keyboard at something
> other than a zs channel (if someone would feel like doing that), but it
> also appears to bring aboard a load of complexity that I think should
> be avoided:
> 
> 	- hardware interrupt hooks for every device
> 	- software interrupt hooks for every device
> 	- both MI and MD front-ends for every device
> 
> I'd like to see the keyboard and mouse drivers be self-contained,
> registering themselves with a capable back-end serving their
> interface (for lack for `streams'), and just relying on an I/O buffer
> in between, and notification running off a software interrupt.

That all sounds very nice.  Let us know when it's done.  (smile)

Gordon