Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!agate.berkeley.edu!cgd
From: cgd@eden.CS.Berkeley.EDU (Chris G. Demetriou)
Newsgroups: comp.os.386bsd.development
Subject: Re: com-beta-0.2.1 now for NetBSD
Date: 13 May 93 13:46:56
Organization: Kernel Hackers 'r' Us
Lines: 58
Message-ID: <CGD.93May13134656@eden.CS.Berkeley.EDU>
References: <CGD.93May12012606@eden.CS.Berkeley.EDU>
<1993May13.190534.3955@cc.umontreal.ca>
NNTP-Posting-Host: eden.cs.berkeley.edu
In-reply-to: surprenc@spp3.jsp.umontreal.ca's message of Thu, 13 May 1993 19:05:34 GMT
In article <1993May13.190534.3955@cc.umontreal.ca> surprenc@spp3.jsp.umontreal.ca (Surprenant Colin) writes:
>I don't think I understand the polemic behind com-beta, did I miss something?
>It look like honest code to me...!
<chuckle> if i hacked it, i can hate it... 8-)
my points about the com-beta code:
(1) the original com driver sucked. it was derived from a hp300
console driver. the interrupt setup on hp300s is
rational, and also the driver had different requirements.
19.2k i/o wasn't one. maybe O(9600) output, human typing
speed input requirements.
(2) the hacks to the serial driver for *speed* while more or less
sound, certainly didn't help the driver's clarity.
(3) my hacks to the driver to add multiport support were non-cute:
the configuration mechanism *sucked*, as did the actual
multiport handling.
(4) my hacks to the driver to do bidirectionality were mostly working,
but generally ill-conceived. to do something like that
properly required a good amount of re-organization
of the driver and/or a *whole* lot more consistency checking
that i was willing to give it. given the fact that the
multiport hacks destroyed the a large part of the structure
of the driver, it was just getting too messy...
<chuckle> perhaps, given all the above, and the lineage of the i[34]86,
that serial driver would be *perfect* for an i[34]86 UN*X, but i'd much
rather see a nicer one...
>what are your plans with NetBSD?
speaking of which, Bruce Evans has a wonderful one, which is eventually
going to creep into NetBSD. (as it is, his npx code is in our current
source tree, his intr code should be coming along eventually, and then
his com code will follow...)
after the com code is in, i understand the Guido van Rooij
(guido@???????.nl) is working on a multiport/bidirectional serial
driver, based on bde's stuff and mine.
honestly, that driver is *very* useful to me. in fact, my machine
at home runs a kernel dated 3/5/93, because no source tree i have
lying around which is dated later than that has had the driver
incorporated, and i *need* the multiport support.
the fact that the driver is is useful, however, doesn't deter me
from wanting it to go away silently; it's really a bad hack,
and a great replacement is on the way...
chris
--
Chris G. Demetriou cgd@cs.berkeley.edu
"386bsd as depth first search: whenever you go to fix something you
find that 3 more things are actually broken." -- Adam Glass