Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.6.2.17 $; site uiucdcsb.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxj!ihnp4!inuxc!pur-ee!uiucdcsb!grunwald
From: grunwald@uiucdcsb.UUCP
Newsgroups: net.unix-wizards
Subject: Re: Wanted: New process scheduler with p
Message-ID: <14900020@uiucdcsb.UUCP>
Date: Thu, 11-Oct-84 17:53:00 EDT
Article-I.D.: uiucdcsb.14900020
Posted: Thu Oct 11 17:53:00 1984
Date-Received: Sat, 13-Oct-84 07:14:02 EDT
References: <9324@watmath.UUCP>
Lines: 19
Nf-ID: #R:watmath:-932400:uiucdcsb:14900020:000:950
Nf-From: uiucdcsb!grunwald    Oct 11 16:53:00 1984


  At the University of Illinois, we use a specific language called Path Pascal
to model O/S behaviour. In the UNIX environment, it currently runs in a single
UNIX process with simulation of multiple processes. Processes are syntactically
similar to procedures in Pascal, and their creation is easy (ie. looks like a
procedure call).

I used this when I took CS323 (our Intro to O/S design class), and I found it
fairly easy to simulate O/S design considerations at a fairly high level. When
I was using it, we were running on a Cyber, which was the *pitz*. It lacked
facilities for separate compilation and inclusion of external "objects" (data
abstractions w/sychronization constraints) was difficult, so it was hard to
be provided with standard "conceptualizations" for disk drives, printers, etc.
Still, it proved useful for modeling paging behaviours, scheduling behaviours,
multi-processor trade-offs, etc.

Dirk Grunwald
University of Illinois