Path: utzoo!censor!geac!torsqnt!hybrid!scifi!bywater!uunet!brunix!cs.brown.edu!lr
From: lr@cs.brown.edu (Luigi Rizzo)
Newsgroups: comp.unix.aix
Subject: PCSIM on RS6000
Message-ID: <63457@brunix.UUCP>
Date: 2 Feb 91 21:52:04 GMT
Sender: news@brunix.UUCP
Reply-To: lr@cs.brown.edu (Luigi Rizzo)
Organization: Univ. of Pisa
Lines: 42

I have an IBM RISC/6000, running the PC simulator, 'pcsim'. In
order for the program to work, an image of the 'C' drive must be
stored in a Unix file (call it C_file). The user can also define
other disks (say 'D', 'E', etc.), that are in fact Unix directories
(call it D_dir).

In order to let several users run their DOS application without
waste of disk space, I would like to use a unique, possibly read-only,
copy of the C_file, containing all the standard programs and files,
and let each user have her/his own D_dir with all the specific
application programs and data files.
However, it looks that this approach is not possible, at
least in a straightforward way, because:

1) the PC simulator creates a lock to signal that a C_file is
   currently used by a simulator process, and refuses to use it if
   it is busy.

2) Programs cannot be run from a D_dir.

I cannot really understand why these restrictions are present.

For 1) I know that there might be problems in having two or more
processes trying to write concurrently to the C_file; however, to
my knowledge there is no restriction in having a read-only C_file
(in fact, you can boot DOS from a write protected floppy disk),
so it would have been smarter to use the lock onlyt to prevent more
than one process to write to the file.

Point 2) is strange too, because executable programs have no need
at all to be treated in a particular way: DOS reads an executable
file the same way it reads a data file, so I see no point in this
restriction.

Any ideas on how to solve the problems ? For 1) it might be easy,
say by creating links to the shared file. Problem 2) might be
much harder, the only solution I see is to use the JOIN command,
but don't know if it suffices. Unfortunately, I've not been yet able
to check the above techniques, so any suggestion will be welcome.

	Luigi Rizzo
	lr@cs.bronw.edu, luigi@sssup2.sssup.it