The biggest issue I see when Stallman views get discussed is the ignorance that without the GNU project, many of the projects people take for granted would never have existed.
A good example is the ongoing replacement of gcc for clang, while forgetting that for several years gcc was the only really usable free C and C++ compiler.
Sorry to burst your bubble, but those of us antediluvians who remember loading sources on nine-track tapes for our Vaxen know of a lot of free software that pre-dated GNU and upon which a lot of the early GNU software was based. For quite a while PCC was _the_ C compiler, and knowledge of the BSD userspace was our shibboleth to distinguish ourselves from the System V heretics.
Open source would have existed without GNU. I would have been different, but it still would have been available and existing in the particular software ecosystem niche that is currently dominated by things with a GNU lineage.
Of course it existed, but then do you remember the ongoing trial between Berkeley and AT&T?
What would have Linus and other have used if GNU wouldn't have been an option during those trials?
Going to buy 386BSD, Coherent and similar?
Do you also remember that most developers only started to care about gcc when Sun decided the SDK was going to be available separately and no longer for free?
USL sued in 91 and it was settled in 93 (was running BSD/386 at home and 4.3 Tahoe at work at the time so I remember the entire affair quite clearly.) Linux at that time wasn't even an option if you wanted a system that actually worked but the period of uncertainty was what allowed it to grow into its current niche. If GNU did not exist there were other options. The fact the GNU tools were the easier path at the time is why were are where we are at now, but let's not pretend that there were not alternatives that would have been equally sufficient if GNU never existed.
Nothing is ever certain when discussing conter-factuals, but if Linux had not had a few years of little competition in the x86 space then yes I think the BSD variants would have become the dominant strain. They were better studied and known in the academic space vs the SysV-influenced core of Linux and the BSD strains have always had better support for the various bits of random hardware out there. Remember that at this time there was no particular assurance that Intel would reach the position of dominance that it currently holds, it was Intel 386 vs Moto 68k vs MIPS with a few Alphas for high-end shops. Linux as an alternative sysv-like kernel with a bsd userland would have worked as well as a world of all *BSD.
I guess it was situational then, but </big_generalization> all top universities in the US had loads of boxes running BSD variants at this time, ranging from big-iron Vaxen down to NeXT cubes and Sun workstations. I am trying to think of a major NSFNet participant that was not producing some open-source BSD-ready software at the time and not many come to mind.
On the other side of the medal LLVM/clang more or less only exists because RMS for a very long time vetoed against a plugin framework in gcc in fear of non-free ones. One could say he took the project hostage for political means. I think it's generally a good thing that RMS takes a rather radical stance on his ideas but sometimes being too stubborn can lead in the opposite direction.
A good example is the ongoing replacement of gcc for clang, while forgetting that for several years gcc was the only really usable free C and C++ compiler.