Date:	Wed, 22 Jul 1998 17:18:27 +0300 (IDT)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To:	Morten Welinder <terra AT diku DOT dk>
cc:	djgpp-workers AT delorie DOT com
Subject:	Re: Ispell and pipes
In-Reply-To:	<199807221318.PAA23691@tyr.diku.dk>
Message-ID:	<Pine.SUN.3.91.980722170813.12232D-100000@is>
MIME-Version:	1.0

On Wed, 22 Jul 1998, Morten Welinder wrote:

> Eli's idea under Windows might be workable.  Using device drivers
> is a bit drastic, but a file version seems doable:

I am actually playing with the file-based idea for some time, precisely
because it can be used as a more-or-less doable prototype.  Armed with
such prototype, we could use it as a testbed for all the basic machinery
that needs to be written to make this a viable solution (all those
functions like `fork', `pipe', `wait', `kill' that can at least send a
SIGINT to the subprocess, and the async subprocesses support in Emacs, for
those who care ;-).  Changing the low-level machinery of how the
interprocess communications work should be relatively easy after that, if
we find the file-based solution unsatisfactory. 

The main advantage of using Windows multiprocessing is (IMHO) that this
aspect is one of the most tested and debugged parts of Windows.  For
example, I routinely run several DJGPP applications at once, each one of
which needs many megabytes of RAM, and I have yet to see Windows crash. 
After all, the first versions of Windows were nothing more than a
protected-mode DOS multi-tasker.  The GUI and all the rest was added much
later, and is much more crappy (again IMHO). 

> Pipes aren't bi-directional -- but programs like ispell will be
> connected using two.

Sure, that's what I meant.  We need a working `pipe' library function.

- Raw text -