On Sun, Jul 27, 2008 at 9:27 PM, edmund ronald <edmundronald@xxxxxxxxx> wrote: > Fréderic, Hal, Nicolas, > > Thank you for taking the time to educate me. > > I agree that this solution doesn't scale or maintain well in the general > case. > As distribution maintainers you certainly have my respect. > Now let me make the instrument-side problem clearer: > > Instrument makers don't necessarily want invisible bug fixes. Say I > have an instrument and I run acceptance tests on it. What I am > guaranteeing to the customer is that all the instruments of this > function match. If there is a bugfix somewhere in the chain, (or > worse a BUG!) that creeps in via the shared library mechanism, then > the semantics of the library change, and suddenly various instruments > on different systems with and without the fix may not match anymore. > > How can I avoid fixes to shared libraries affecting the behavior of > an instrument ? Making sure the needed fixes are pushed down the stack to "distributions" is the best way, IMO, because those fixes will not only help one particular instrument but improve the entire stack. Again, it is a sense of eco-system. Linux (and other free software OS) can be improved in such a way, in the long run, those individual fixes would become useless. Of course, doing this work is not always very fun, it requires more time and energy than just static linking, but I think it is not wasted energy. -- Frederic Crozat