From:	Charles Sandmann <sandmann AT new-orleans DOT NeoSoft DOT com>
Subject:	Re: QEMM's VM manager, V2, and CWSDpmi
To:	turnbull AT shako DOT sk DOT tsukuba DOT ac DOT jp (Stephen Turnbull)
Date:	Wed, 21 Dec 1994 09:16:18 -0600 (CST)
Cc:	RGRUNWAL AT wasp DOT cs DOT cowan DOT edu DOT au, djgpp AT sun DOT soe DOT clarkson DOT edu

I hope to take a look at the QDPMI virtual memory issue soon.  If I can't
find a workaround, getting CWSDPMI in top shape becomes an even higher
priority.

CWSDPMI is currently based on modified GO32 code and built with TCC/TASM.
It will be a full 32-bit-only 0.9 implementation when it is released, probably 
before V2.0 is released.  CWSDPMI can be used for some applications other than
DJGPP (with RSX, for example, a different GCC DPMI-only port for DOS).  It is 
hoped that the 0.9 release will be good enough until the 1.0 release is
finished.  The current beta is missing HW interrupt support, but seems
to have everything else working.  It wouldn't have been possible without help
from Morten Welinder.  I expect it to work inside of DVX (but probably not
if loaded before).

My plans are to pretty much abandon this code as a prototype for the 1.0
DPMI spec release.  The next version should allow mixing 16-bit and 32-bit
DPMI clients (so you can mix MSC/BCC tools with DJGPP apps).  It will be
written in DJASM (a free 16-bit assembler which comes with V2) and GCC/GAS,
and will have small memory footprint.  Realistically, I don't think the 
1.0 version will be complete until 4Q95 or later.

BTW, one of the byproducts of this work is a DPMI-only GO32, which only
needs around 50K of memory per nested program, down from the 130K 
which GO32 currently needs.  This allows V1.1x programs to run nested
10+ deep.  This is definitely throw away code, but if anyone wants the 
source+binary for this to tide you over until V2 is out, contact me
directly.

- Raw text -