I just found a reference to Diego Pettenò's article about gnulib: http://www.linux.com/archive/articles/50702
Its summary section is quite interesting, because despite being written six years ago, it still lists what I see as the main defects of gnulib (licensing, dependency on autotools, source library). Of those three defects, the third is now being addressed (at least, for POSIX APIs, by libposix), and the second is not particularly urgent (autotools has a much bigger reach than it had 6 years ago, and has less competition; moreover, by using libposix one doesn't need to use autotools in a dependent package, at least for POSIX APIs, ignoring the rest of gnulib's goodness). However, licensing is a problem. In particular, I'd love to use gnulib in other projects on which I work, such as file (Christos Zoulas's version, used widely in BSD and GNU), but it seems I can't, because it's BSD-licensed. I assume the licensing for gnulib arises from standard GNU policy; I just wonder if the portability parts may be a case for an exception. I imagine fondly being able not only to simplify and solidify file (which is a bit of a dog's breakfast in the same way as much GNU code was pre-gnulib), but also of the opportunities, in both directions, for getting portability code out of programs such as OpenSSH and into gnulib. gnulib seems to have made a big difference wherever it's been used (it certainly has to me), and yet only a fraction of its potential uses have been tapped...and it seems to me that there's a huge opportunity for the entire free software ecosystem here lying unexploited. -- http://rrt.sc3d.org
