[1082] in SIPB_Linux_Development
Re: ELF migration. (Go lemmings go!)
daemon@ATHENA.MIT.EDU (Theodore Ts'o)
Thu Oct 5 22:26:14 1995
Date: Thu, 5 Oct 1995 22:25:54 -0400
From: "Theodore Ts'o" <tytso@MIT.EDU>
To: Jered J Floyd <jered@MIT.EDU>
Cc: Derek Atkins <warlord@MIT.EDU>, eichin@MIT.EDU, jered@MIT.EDU,
linux-dev@MIT.EDU
In-Reply-To: Jered J Floyd's message of Thu, 05 Oct 1995 22:19:46 -0400,
<199510060219.WAA16152@vorlon.mit.edu>
Date: Thu, 05 Oct 1995 22:19:46 -0400
From: Jered J Floyd <jered@MIT.EDU>
Why do you think it is necessary to have a seperate machine as an a.out
build engine? If properly configured, an ELF machine can build a.out
executables, and we shouldn't be planning on building many new a.out
binaries anyways.
This is not completely true. At the very least, it is impossible to
build new DLL shared libraries once you've converted over to ELF. It's
one of the reasons why I haven't converted; I still want to be able to
generate the libext2 DLL shared libraries easily. On the other hand,
SIPB probably doesn't need to build any DLL shared libraries, so this
might not be a problem.
I do agree with the previously expressed sentiment that most important
thing we can do is create easy-to-install packages that allow people
with a.out systems run ELF binaries. Once we do that, if people want to
start using ELF, we might as well. *Someone* is going to have to flush
the bugs out of the 5.x libc. :-)
- Ted
