Date: Sun, 26 Jul 1998 11:52:02 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Charles Sandmann <sandmann AT clio DOT rice DOT edu>
cc: salvador AT inti DOT gov DOT ar, djgpp-workers AT delorie DOT com
Subject: Re: Ispell and pipes
In-Reply-To: <9807231333.AA16334@clio.rice.edu>
Message-ID: <Pine.SUN.3.91.980726115140.26387B-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Thu, 23 Jul 1998, Charles Sandmann wrote:

> You want to avoid 0x21 and 0x31 since they make things slow...

Could you elaborate about why Int 31h makes it slow?  

It does seem like an ideal candidate for DJGPP programs.  We only need
to watch for a single DPMI function 0x0300 (at least in the first
approximation), and there's no mode switch involved, or so it seems...

If I understand correctly, installing a PM hook for Int 31h does NOT
require it to be locked, so the _go32_XXX helper functions can be used
without fear.  Is that correct?

> The DPMI server only knows about one thing - how deeply nested it is.

How about memory allocation: doesn't the DPMI host need to know to
which client does the memory belong?  Otherwise, how would it release
that memory correctly when a nested client exits?

The same question applies to allocating selectors (or any other
resource, for that matter).