Date: Mon, 14 Oct 1996 07:43:07 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Mark Habersack <grendel AT ananke DOT amu DOT edu DOT pl>
Cc: "John M. Aldrich" <fighteer AT cs DOT com>, djgpp-workers AT delorie DOT com
Subject: Re: Install thingy
In-Reply-To: <Pine.NEB.3.95.961013202854.5686D-100000@ananke.amu.edu.pl>
Message-Id: <Pine.SUN.3.91.961014073641.16204R-100000@is>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII


On Sun, 13 Oct 1996, Mark Habersack wrote:

> On Sun, 13 Oct 1996, Eli Zaretskii wrote:
> 
> >I urge you to reconsider.  Hard as it might be, writing it in DJGPP 
>
> It's not a matter of hardness.  I was rather thinking that the install utility
> HAS TO run on any system. Doesn't matter whether it's a machine with 8088 or
> Pentium.  Install should run gracefully on as small as 256KB of conventional
> memory (I've seen some high-end computers with that much free!!).

IMHO, install need not run on anything that can't run DJGPP.  Detecting 
such machines is the task of djverify; no need to duplicate efforts and 
code.  Run djverify at the beginning and if it returns one of the fatal 
errors, gracefully exit telling user to upgrade.

If djverify detects non-fatal problems (like missing environment 
variables and such), you can repair those problems temporarily (by e.g. 
setting the missing variables) and do the installation.

If the problem that bugs you is the chicken-and-egg problem, namely that 
you need to install DJGPP in order to run a DJGPP program (the 
installer), then it can be easily solved: all you really need is a DPMI 
host.  So having a stand-alone image (with PMODE or Charles unreleased 
version which writes CWSDPMI image the first time it runs) is all you 
need to do.