Date: Wed, 22 Jul 1998 15:18:01 +0200 (METDST)
Message-Id: <199807221318.PAA23691@tyr.diku.dk>
From: Morten Welinder <terra AT diku DOT dk>
To: djgpp-workers AT delorie DOT com
In-reply-to: <m0yydcV-000S4XC@inti.gov.ar> (salvador@inti.gov.ar)
Subject: Re: Ispell and pipes
References:  <m0yydcV-000S4XC AT inti DOT gov DOT ar>
Precedence: bulk


It's a neat idea which, I think, has occurred to many of us over
the past few years.  My own, personal conclusion has been "where
is the fire exit?"

As Charles will tell you, the various DPMI providers -- including
the Windows ones -- will deviate from the standard in interesting
ways and will contain bugs that only bite you once every so often.
The more so for thing -- like RMCBs -- that are not too well tested
in the field.

Moreover, the DPMI 0.9 spec. has a number of gotchas with respect
to running more than one client at a time.  Consider two programs
taking over the 0x1B vector, for example.  They will share the
same IDT and the last to start must be the first to exit, or else!

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

  * The reading side waits for file "p-nnnnn.in".  When it appears,
    it is read and deleted.
  * The writing side writes to file "p-nnnnn.out".  At suitable times,
    when the *.in file isn't there already, the file is closed and
    renamed to *.in

Pipes aren't bi-directional -- but programs like ispell will be
connected using two.  That should work ok, I think.

Morten