Subject: Re: HEADS UP: migration to fully dynamic linked "base" system
To: Jason R Thorpe <>
From: Johnny Billquist <>
List: current-users
Date: 08/27/2002 03:29:14
On Mon, 26 Aug 2002, Jason R Thorpe wrote:

> On Tue, Aug 27, 2002 at 10:05:57AM +0900, wrote:
>  > 	in other words: if the motive is locale support in binaries in /sbin,
>  > 	we should solve it by providing dlopen() and friends to statically-
>  > 	linked binaries.  solving it by avoiding statically-linked binary looks
>  > 	to me a wrong way (caveat: i'm no dynamic linkage guru).
> As has been pointed out a few times in this thread, adding dynamic
> support to static binaries is extraordinarily difficult (if not
> impossible, depending on how much you're willing to fudge the
> semantics).

Well, let's be honest here.
It's not that difficult, but it's a problematic thing, since we'll have
potentially conflictiong library versions running, and no way of detecting
If we're willing to accept this possible lossage in statically linked
binaries we just need to get dlopen() and friends loaded in static
binaries, and export the symbols. I can't imagine that being *so* hard.

But all hell might break loose if some library changed semantics or the
interface to a function that had been statically linked into a program.

We'll have possible security issues, possible system corruption, and god
knows what else.
I guess a requirement would be that statically linked programs would be
required to be rebuilt when shared libraries are rebuilt.
(Hmmm, no, that won't be enough. We need to rebuild everything that
statical binaries refer to, and then rebuild those binaries.)


Johnny Billquist                  || "I'm on a bus
                                  ||  on a psychedelic trip
email:           ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol