Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site cadre.ARPA
Path: utzoo!linus!philabs!cmcl2!seismo!rochester!cmu-cs-pt!cadre!km
From: km@cadre.ARPA (Ken Mitchum)
Newsgroups: net.micro.pc
Subject: Re: JOVE/MSDOS question
Message-ID: <446@cadre.ARPA>
Date: Wed, 5-Jun-85 17:33:41 EDT
Article-I.D.: cadre.446
Posted: Wed Jun  5 17:33:41 1985
Date-Received: Sun, 9-Jun-85 22:26:07 EDT
References: <803@mako.UUCP>
Reply-To: km@cadre.ARPA (Ken Mitchum)
Distribution: net
Organization: Decision Systems Lab., University of Pittsburgh
Lines: 46



>Has anyone looked over the sources distributed by
>Ken Mitchum for the MSDOS version of JOVE?  One of
>the problems with this version is that on a floppy
>system the swapping to a temporary file on disk will
>kill you.
>
>It appears after looking at ``temp.c'' that a provision
>has been made for doing in-memory buffering instead,
>but only for the VMUNIX version.  Simply taking out the
>#ifdef should allow this to be done for MSDOS systems as
>well.  Of course, you need lots of memory to do this.
>Has anyone tried this?
>
>	Glen McCluskey
>	..tektronix!mako!glenm


Yes - swapping on a floppy-based system is very slow. This is mentioned
in the documentation - however, I plead completely guilty to not looking
at the problem before. I worked on JOVE on hard disk systems only, and
only tried it on a floppy system briefly before distributing it. I also
assumed most people would use Jove on a hard disk system, since just
compiliing and linking all the files on a floppy system would take forever!

I have never looked closely at the code in "temp.c" as it required no
modification to work with MSDOS (sqeaky wheel gets the grease). I will
try to take a look at the code and see how in-memory buffering works
out. To use more memory with C86, it is necessary to redefine _MAXFMEM
and _MINFMEM, which were not adjusted in the JOVE sources, so that a
version would work with a minimum of memory (256K). My schedule is very
busy at the moment, but I will post something to the net regarding this
in the future, along with any code changes. Alternatively, any one else
implementing in-memory buffering should post comments on their experiences.

There are MANY problems with JOVE/MSDOS. It should be taken as a kludge
which works, which is the spirit in which I worked on it. There are many
optimizations and changes to be done, particularly with how the screen
images are handled. Alas, I wish I had more time to spend on it.

  Ken Mitchum
  University of Pittsburgh