On Tue, Sep 13, 2005, Birger Krägelin wrote:
> > So, the main argument against shared libraries is the support
> > of multiple instances, currently.
>
> What's the argument against static linking?
> Memory shouldn't be a problem anymore...
Yes, neither HD nor RAM consumption is the real problem today (HD is
cheap and RAM is not consumed very much more thanks to smart enough VM
management in modern Unix flavors based on copy-on-write semantics,
etc).
But one problem with static linking is that in case of security issues
(think "zlib" or the recent "pcre" issues) one is required to rebuild
lots of applications. Nothing one cannot handle, but it at least
requires lots of more efforts.
The second issue (which I guess triggered Matthias here ;-) is that
sometimes (think "gtk" here) the static linking is rather hard to
achieve because the upstream authors sometimes seem to no longer test
it at all and we are required to hack their code a lot. With a shared
library approach we could avoid this hacking.
These are mainly the two major issues AFAIK which trigger the requests
for shared libraries. My personal opinion since a longer time is
that if (AND ONLY IF) we are able to OPTIONALLY and LESS INTRUSTIVE
integrate the support for using shared libraries, the advantages (those
two issues) outweight the disadvantages (the intrusive hacking, the
portability issues, the less robust and slower run-time, etc). Until
know I was not able to imagine myself how we can achieve this, so for me
I currently see more disadvantages than advantages. Hence I personally
still prefer to stick with the static linking, although I'm certainly a
Dynamic Shared Object (DSO) fan (I worked a lot on the DSO facility for
Apache many years ago...)
Ralf S. Engelschall
[EMAIL PROTECTED]
www.engelschall.com
______________________________________________________________________
The OpenPKG Project www.openpkg.org
Developer Communication List openpkg-dev@openpkg.org
