I am not arguing against any of that. I love the transparency as much as you do and I always used it as one of the stronger pro-BeOS arguments when evangelizing.
I'm also in the pro-transparency camp. Please, don't hide any details of the system from me, even if they're not that pretty (in case of GNU-based software). It's the only way to learn the ins and outs of a system.
My point is that we cannot do completely without any ports.
Make that "many ports" and I'd agree. ;-)
These may require certain (scattered?) ways of being installed. And even for native software, we may run into dependency and binary compatibility issues sooner or later. These problems need a solid, reliable solution.
I don't think you can really solve these problems, except at the source: write unbroken software and libraries. Remember that the source of the problem is buggy library development. Any "solution" will just be a workaround; and whatever you do, there will always be cases when the designed system will break. It really is a question of "lesser evil" than of "good solution".
Once integrated search for and installation of software, automatic updates and these things come into play, some form of smart package management backend seems to be inevitable.
Yes, but faced with the variety of problems, it is questionable if one can write a PM system which is smart enough to reliably solve all problems. Such a PM may also turn out to stand in the way of the advanced user, and I'd hate this to happen. I'd rather rely on the most advanced package management system known to man - man. Christian