Static linking considered harmful

Axel Thimm Axel.Thimm at ATrpms.net
Fri Nov 24 19:46:33 UTC 2006


On Fri, Nov 24, 2006 at 08:28:54PM +0100, Arjan van de Ven wrote:
> Maybe the real answer is something entirely different: if it was easy to
> automatically package up a binary and all the libs it needs into
> something that's self executable/self contained... that would take care
> of the requirement too. Maybe some answer is the real thing

you mean http://statifier.sourceforge.net/?

> However I think it's very important to separate out "we want portable"
> and "we require static linking", since just about ALL arguments for
> static linking actually were not about static linking but about wanting
> portable.

That kind of argument separation and liquidation is an eloquent
speaker's trick. In that sense there are no arguments in favour of
dynamic linking either, Ulrich wants to have security and
maintainability, he doesn't really care about dynamic linking, right? ;)

FWIW again from the number crunching community, sometimes statically
linking numerical libs shows performance gains (although when the
problem domain is of that kind the libs tend to be headers-only with
inlining), and some only commercial available libs only offer static
libs.

Which brings yet another argument in favour of not disallowing
statically builds: ISVs love to use these in order to have one build
for the whole Linux world. Not that these builds are ever a joy to the
users, and some may say the noone cares about proprietary browser
plugins or java vms, but I would be sad to not have a Linux way to
partition HP storage for example.
-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/fedora-devel-list/attachments/20061124/f64d0908/attachment.sig>


More information about the fedora-devel-list mailing list