From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10108092059.AA12546@clio.rice.edu>
Subject: Re: gettext port (DXE prototype)
To: jeffw AT darwin DOT sfbr DOT org
Date: Thu, 9 Aug 2001 15:59:51 -0500 (CDT)
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <20010809103734.A13133@kendall.sfbr.org> from "JT Williams" at Aug 09, 2001 10:37:34 AM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> Can you say if/why the ready-made dxe implementations in v2tk/ are not
> satisfactory for this purpose?  (I ask because I have been experimenting
> with v2tk/dxe2_011.zip, but perhaps it has limitations I am not aware of.)

I planned to look at the enhanced DXEs after taking a quick stab at it with
the standard tool.  The example had every combination that was interesting:
it needed to export routines and data structures, it needed to import
routines and data structures (abort() and errno).  The number of total
values in the transfer vector was small.  After writing DXEs I always
planned to show such a real example and never got the chance.  It also
showed me that a switch I had in my local dxegen code from about 6
years ago never made it into the DJGPP tree (ability to have DXEGEN
create an export list and an import list).  Sigh.  

Don't take this effort as any rejection or critque of the other dxe/dll
tools - look at it as an example or prototype used on a real world problem.
Proof of concept (you can write full scale DLLs using DXEs if you spend
enough effort).  The lightweight approach (total new code added < 50
lines, rest already linked in).  A baseline.

I believe in prototypes - trying things 2 or 3 ways to see which is best,
then save the best and toss the rest into the box for future prototypes.
I plan to look at the other available dynamic loading tools to see what
nice features/good ideas exist.

The final result might be shipping images with a separate loadable section,
maybe not (maybe too much trouble).  Maybe it's built with DXE2.  Or some
other DLL tool.  Maybe some additional backward compatible features on
DXE.  (Example: write available import/exports.  transfer vector 
definition - default based on imports/exports.  "glue" code generated
based on vector definition).

Ideally the best outcome is use and endorsement of some dynamic loading
tool which is so easy that dynamic loading becomes common where it makes
sense.

So in a phrase - I haven't gotten around to it :-)