Thanks to Lee for pointing out that through his, mine, and other efforts gentoo packages will remain up to date and available for development and hopefully production use.
I, too, face the problem up keeping a work Redhat box up to date with the possibility of making a hard decision between Fedora, Enterprise ($$) Redhat, or Other distro next spring. I think we will end up going with one flavor of RH or another (either Fedora like I said or Enterprise).
I think Debian has shown how easy the administration of a distro *can* be.
I think Gentoo has shown how easy to configure, understand/learn, and keep extremely up to date a distro *can* be. This is my home development and deployment distro.
But no one to my satisfaction has solved the Really Big problems, some of which have been mentioned in this thread: API upgrade paths to keep binary and API compatibility over several years on a box. Java and .Net are designed with that more in mind, and I'm sure that WinLonghorn will be closer to it than Win32 is for sure.
The real winner will probably be DragonFly BSD, which is spinning off of FreeBSD 4.x (and something I may move to next after it gets some momentum):
http://www.dragonflybsd.org/Goals/packages.cgi
This is one of the most exciting and concise descriptions of the problem and proposed plan I have come across--highly recommended if still theoretical reading.
