X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
NNTP-Posting-Date:	Mon, 26 May 2014 23:42:49 -0500
From:	"Charles Sandmann" <cwsdpmi AT earthlink DOT net>
Newsgroups:	comp.os.msdos.djgpp
References:	<op DOT xfpi3cic6zenlw AT localhost> <201405111842 DOT s4BIgrRx012234 AT delorie DOT com> <op DOT xfprrlvs6zenlw AT localhost> <PKedneUwMOJcSe_OnZ2dnUVZ_u2dnZ2d AT earthlink DOT com> <op DOT xft25kb26zenlw AT localhost>
In-Reply-To:	<op.xft25kb26zenlw@localhost>
Subject:	Re: secret DJGPP documents?
Date:	Mon, 26 May 2014 23:42:47 -0500
MIME-Version:	1.0
X-Priority:	3
X-MSMail-Priority:	Normal
Importance:	Normal
X-Newsreader:	Microsoft Windows Live Mail 16.4.3528.331
X-MimeOLE:	Produced By Microsoft MimeOLE V16.4.3528.331
Message-ID:	<IKydnTcfp49UihnOnZ2dnUVZ_vednZ2d@earthlink.com>
Lines:	53
X-Usenet-Provider:	http://www.giganews.com
NNTP-Posting-Host:	64.91.136.27
X-Trace:	sv3-fsfBgj238crZHYWYx/qLI3sw49vO65eSDkGTdyWwCzhC4VuOfqeZLUWNBwMhjryyfBcEOYg2ESZVMcy!grNPvSNHqbWG3fsXINLfOTu0Ct21VE77cRhNwpPhKB0hUfrbvA0KERrqcGZ2iTNgBfrKyhJzYlJC!97GfeGsAesFf5jkbOPB9pSfiuJ9NegU=
X-Abuse-and-DMCA-Info:	Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info:	Otherwise we will be unable to process your complaint properly
X-Postfilter:	1.3.40
Bytes:	4315
X-Original-Bytes:	4254
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

>"Rod Pemberton"  wrote in message news:op DOT xft25kb26zenlw AT localhost... 

>I'm sure you've probably answered this before, but why was a
>replacement (MWDPMI) for CWSDPMI planned originally?

CWSDPMI required a commercial compiler TurboC (which wasn't free)
to build.  Later I started building with BorlandC (also not free) when I 
wrote my own libc routines.  The resulting code is not very tight
(not a lot of optimization happens without the developer rewriting
code to make it happen).  It also had some areas of the code which 
could be cleaned up.  The goal was to build with a GNU provided
toolchain.  Eventually Borland made TurboC free so users could build
it themselves.  Windows provides built-in DPMI.  So the need
decreased while the authors' time became less available.

When DJGPP v2 was in development, we had concerns about what
DOS users would do.  Stick with v1?  Be forced to buy commercial
DPMI to run the free compiler and executables?  The original plan
was always to have DPMI be fully built with the toolchain (using
DJASM and GCC).   But that was a big project at the same time we
were trying to get v2 feature complete and bug free.  So I did a 
quick modification of DJ's GO32 and got a very minimal DPMI 
server working in a few weeks of work.  Go compare the GO32
source to CWSDPMI sometime.  So CWSDPMI was always
just another week or so of effort before it would be frozen as
good enough and we would work on the next generation.   There
never were a lot of CWSDPMI releases - r1 to r3 were in 1996,
r4 in 1998 ... and the rest was just minor updates and big 
memory handling.  You can probably check the archives for the
number of times I said this will be the last release of CWSDPMI.
I'm glad other people have had the time and motivation to
provide alternatives so I don't have to worry about it.  I still 
hear about distributions using older versions (like r3). 

Never underestimate the effort to bring something to production
that talks at a low level to the hardware.  There were issues with
some types of machines, or hardware interrupts, that took 
months to track down and debug with CWSDPMI.  In the early
releases there were lots of people testing with simple test 
programs.  Still today occasionally I hear about new hardware
that doesn't handle A20 quite right (who cares anymore?)  But
there just wasn't the time (or drive) to finish MWDPMI.  I
still have an ancient Compaq laptop that DPMIOne will hard 
hang in some configurations.

CWSDPMI had a big installed base, so commercial companies
also helped out - like Symantec sending me a patches to work
around bugs in Intel chips.  They got Intel to help them figure 
that out ... things dealing with instruction sequences.  I also 
got feedback from id on the Quake user issues early during
the betas.

- Raw text -