> Neither Linus nor DEC say they are working on shared libraries themselves.

Linus does.

This is a related posting I saved:

From: torvalds@cc.Helsinki.FI (Linus Torvalds)
Newsgroups: comp.unix.bsd.netbsd.misc,comp.os.linux.advocacy
Subject: Re: DEBATE:  BSD vs. Linux
Date: 6 Sep 1995 19:52:00 +0300
Organization: University of Helsinki
Lines: 27
Message-ID: <42kjjg$4u@klaava.helsinki.fi>
References: <4233kp$t8p@hilly.apci.net> <42gn16$98f@klaava.helsinki.fi> <MICHAELV.95Sep5002229@MindBender.HeadCandy.com> <42hhvh$h5g@news.cloud9.net>
NNTP-Posting-Host: klaava.helsinki.fi
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit

In article <42hhvh$h5g@news.cloud9.net>,
Thor Lancelot Simon <tls@cloud9.net> wrote:
>>
>>Just for the record, the answer to your "?" is yes, NetBSD/Alpha runs
>>OSF/1 binaries (that's how it was initially bootstrapped).  It's just
>>one of a vast collection of well-abstracted compat code, actually.
>
>It is worth note that NetBSD/Alpha does _not_ currently run dynamically 
>linked OSF/1 binaries due to undocumented aspects of the behaviour of the
>OSF/1 shared library loader, unless cgd has made a big change since I 
>last looked.
>
>Perhaps the Linux code could be used as documentation. :-)

You're quite welcome to.  But _real_ men do it the hard way.. 

Seriously, the way I did this was by using a special /sbin/loader binary
with debugging hooks that I made ("dd" is your friend: binary editors
are for wimps) and taking a look at what the stack etc environment
looked like at process startup.  As a result the linux sources have no
explanations in them except for the single comment "/* whee.. 
test-programs are so much fun.  */", but they do seem to work. 

(I tried to get some documentation out of Digital on this, but as far as
I can tell even _they_ don't have it ;-)

		Linus