Mike Hearn's 'Linux Problems' page has some interesting assertions on how to 'fix' Linux and make it 'enterprise-supportable' (which seem to boil down to: distributions suck, so avoid them; Python sucks, so avoid it; C++ sucks, so avoid that too; glibc, gcc and headers suck, so they need fixing; shared libraries, too, are hard for us, so we'd like other people to avoid dealing with them because autopackage is king).

The Debian bit is just flat-out wrong:

Debians dependency scanner is broken anyway, as there's no way to tell what version of a library is actually needed. Only the developers know that. It works around that problem by assuming that the binary requires whatever version it was built against (which as we see elsewhere may be true or may not be).

Cute, but no. Packagers can specify in shlibs files, the last ABI break. So if your library adds a new interface in 2.1.3-1, then you can put:
libfoo 1 libfoo1 (>= 2.1.3-1)
in your shlibs file, and wham, anything built against libfoo now gets that dependency. Of course, if the downstream packages are really ninjastupid, they can hardcode in their library deps.

Why should the convenience of Debian packagers, who number in the low 2s, outweigh the convenience of people who download autopackaged binaries from the website, who number in the high 200s/day?

Actually, last I checked, there were close on 1000 registered Debian developers, who had gone through the long process. I don't know how long the New Maintainer queue is (except: really long). There are millions and millions of users of Debian and its derivatives (Ubuntu, Knoppix, Maemo, et al). If you're going to be a condescending twat, could you at least please try to be accurate, instead of picking on Debian for being insignificant compared to your hundreds of downloads per day?