Xref: utzoo unix-pc.general:3935 comp.sys.att:7857
Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!att!icus!limbic!gil
From: gil@limbic.UUCP (Gil Kloepfer Jr.)
Newsgroups: unix-pc.general,comp.sys.att
Subject: Re: SysVR3 port for UNIX PC (*LONG*)
Summary: Size of kernel
Message-ID: <573@limbic.UUCP>
Date: 22 Oct 89 06:47:38 GMT
References: <2917@umn-d-ub.D.UMN.EDU> <987@icus.islp.ny.us> <2928@umn-d-ub.D.UMN.EDU> <990@icus.islp.ny.us>
Reply-To: gil@limbic.UUCP (Gil Kloepfer Jr.)
Distribution: na
Organization: ICUS Software Systems, Islip, NY
Lines: 48

In article <990@icus.islp.ny.us> lenny@icus.islp.ny.us (Lenny Tropiano) writes:
>In article <2928@umn-d-ub.D.UMN.EDU> rhealey@ub.d.umn.edu (Rob Healey) writes:
>|>	named Intel... Appologys to Lenny, but do we lose the
>|>	possiblity of a SysVR3 UNIX because it can't support some of
>|>	the fringe areas of the UNIX PC? I could live with having to
>Ok, want some reasoning!!  I'll give you some :-)

I'll add some to Lenny's article and try to calm the small flames I'm
starting to see eminating from this conversation (maybe I'll make some
new ones ;-).

I'll start by saying that this was all beaten to death at the UNIX-pc
BOF at Usenix, but there is one serious consideration that we can't
get around and that's the 4MB address space limitation in the memory
management hardware.  I'm sure that the motherboard could be hacked up
to make it work, but I can imagine it would be a royal pain.

After pondering with many people in many circumstances over the issues
of existing software, memory management hardware limitations, device
driver rewrite issues, and loss of some "neat" features in our kernel,
it is *my* opinion that we'd be better off cleaning up the UNIX-pc
kernel and maybe making it more complient with SVR3 than to attempt a
port of SVR3 to the UNIX-pc.  Again, with stress on the memory management
problems, as one eats up more space with a kernel, one has less space to
work in.  Assuming the port could be done and drivers and software
working in some kind of compatibility mode, we would still be left
with a problem of a kernel of a larger size than the current one, almost
guaranteed.

I certainly don't say that those interested in performing a port should
not do so, and I think constructive discussion about a port would be
good.  I am saying that for all that is gained by porting SVR3 to the
UNIX-pc, you may end up losing some more important resources of
the machine.  Further, if you _really_ need a machine that can support
SVR3, buy a 386 (choke, gag ;-).

Lastly, and I think a discussion that also has merit (although there
are those who would disagree) is a port of some version of BSD to the
UNIX-pc.  This was also discussed at the BOF, and it seemed to have
more of a following than a SVR3 port.  Again, followers take note--
make sure that the size of the kernel is less than or equal to the
size of the current one or you'll end up with applications that don't
run.

------
| Gil Kloepfer, Jr.
| ICUS Software Systems/Bowne Management Systems (depending on where I am)
| ...ames!limbic!gil