hmm, i felt compelled to create an account specifically to reply to this thread, as i have been considering something similar for my own software endeavors.
the bottom line: it requires an experienced developer/team many person-months to create a quality product; time=money. the feel i get after reading 5 pages worth of comments, is that few have written a line of code in any language, let alone professionally for a company or themselves. if you want the best, latest, and absolute greatest that a development team has to offer, then offer some compensation for their energy. isn't that fair? don't we exchange money for everything else in our lives? software isn't any different.
i am an avid open source/FOSS supporter, and have used nearly 100% open software for many years; i enjoy spreading the good word to others. however, i believe in the rights of a person/group to assert ownership over something they've created.
FOSS (in the modern economic/social/governance paradigm) simply doesn't fit very well for certain categories of software, especially direct, end-user consumables. it works better for underlying, shared, component-like technologies such as libraries, platforms, protocols, etc. etc... anything that isn't as directly marketable and can be shared and disclosed across many companies with little impact on their individual top level products. grasp the fact that companies/developers need to recoup capital lost in development. this is why Apple has contributed to LLVM/Clang.
take a company like NoMachine. they produce an awesome library for efficient X server remote displays, like thin clients. the library is GPL i believe, but their IMPLEMENTATION using the library + additional features (their closed-source server) is not free. they use the money generated by the server product to fuel development of the library... which is now used by open projects like FreeNX and NeatX, both of which would not survive without the NoMachine's continued development of the underlying library.
i have developed professionally for several years, and am now a self employed developer. i am considering a similar process for an upcoming product of mine: the core product will be open, but some of the heavy lifting modules that do the really interesting stuff will not be open... at first anyway. similar to the author, i will release the modules after i've had a chance to recoup the costs by licencing it to companies.
in case no one bothered to read it, Sprewell posted a good link about the ghostscript author:
there are several ways for closed and open source software to work TOGETHER, to create higher quality products on both sides of the fence, without being greedy and still being fair to all parties. a purely FOSS software industry is an unimplementable pipedream.