[zem]

one of the most striking examples of this is rms's objection to a modular gcc. modularity is a technically superior solution, and would have been supremely useful in supporting a larger ecosystem of development and analysis tools (e.g. ides, automated refactoring, source-to-source translators, etc), but would also have made it easy to build proprietary systems that could remain closed while using modules from gcc.

the current groundswell of clang adoption has revealed just how many people wanted a good, free, modular c++ compiler, but by rms's principles, making gcc serve their purposes would have come at too high a cost.