From: elf AT netcom DOT com (Marc Singer)
Message-Id: <199605141933.MAA22597@netcom10.netcom.com>
Subject: Re: ELF wanted
To: martynas DOT kunigelis AT vm DOT ktu DOT lt (Martynas Kunigelis)
Date: Tue, 14 May 1996 12:33:56 -0700 (PDT)
Cc: djgpp AT delorie DOT com (DJGPP List Alias)
In-Reply-To: <199605141156.HAA12461@delorie.com> from "Martynas Kunigelis" at May 14, 96 02:42:54 pm
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 2294      

> Well, now that we have complete ELF documentation, someone (e.g. me)
>   could > start moving DJGPP to it. But here again arises the
>   rethorical question: IS > IT WORTH IT? With ELF, we could use more
>   Linux binary tools etc., and > have the DLL implementation made
>   easier. But what about Win32? AFAIK Win32 > uses COFF, and, as
>   Malcolm said, if someone writes Win32 support (will anyone?  >
>   ?????), we could use the advantages of Win32 COFF-based DLL
>   engine. So the > point is: either we (?) start making the Win32
>   support or begin adopting > DJGPP for ELF. But wait, could it be
>   that both ways were possible? I mean > having Win32 support with
>   ELF? Or, perhaps, move Win32 to ELF ;-) ?? Bill?  > Malcolm? What
>   do you think? Is adding Win32 support *very* complicated? Or are >
>   header files and that kinda stuff the only problem? Is a different
>   linker > reqired or just a different stub loader?

Hmm.  It seems to me that the push for ELF is an academic one.  We
certainly don't NEED it.  Whereas Win32 support has some very
important implications for the future of free software.  I've been
writing Windows applications for ten years with Microsoft tools, and
if I can describe them in a word it would be `suck'.  It takes a 32 MB
machine to develop for Windows 95 using their new compiler v4.0.  

Practically speaking, I'd say that neither is particularly difficult.
We'd have to generate the new exe format to get the programs to load,
but that's probably the hardest part aside from linking to the Windows
import libraries.  I believe that these formats are published.

ELF is a fine idea, but rather irrelevant when it comes to the
DOS/Windows environment.  The utility of it is still somewhat lost on
me even though I think it would be an interesting project.  I suppose
we could implement an ELF loader for Windows, but what is the point
when the OS [sic] doesn't recognize the format?

Perhaps the better course is to work with binutils to create Windows
compatible EXEs containing stabs or COFF debug info.  Would it not be
illuminating to recompile Microsoft's MFC libraries with GCC and
compare code size and performance?  

> Enough for now, it's WAY too hot in Lithuania right now...    :)

Do you mean to say that you live there?