I don't want to advocate for a more "pure" definition of open source. You can widen or shrink the scope of anything at any time. I am perfectly OK with a vague

I don't want to advocate for a more "pure" definition of open source. You can widen or shrink the scope of anything at any time. I am perfectly OK with a vague definition of open source that encompasses hardware too. But that is not an opensource definition that I personally (and I would argue most) care about all that much. Few people, if any, are interested in modifying the firmware inside their microwave or refrigerator. Sure, at some point such firmware could become burdensome (e.g. coffee maker with DRM'ed k-cups) to the point where opensourcing the firmware would be something I could get behind. For you that tipping point is different. Now coming back to the example, that is a great example. I think we can learn a lot from that. One thing is that 'open' is a large cafeteria. You would agree that simply opening the source is not sufficient in most cases. If the spec for EXT2 changed after every minor update to the kernel that would make your addition pointless. Same thing with the API at the I/O level. If that changed all the time, it would make your addition brittle and prone to breakage. So what is it that's really beneficial here? I'd argue that the source is the least important here. Open protocols, open data formats, defined interface boundaries, and the like are _much_ more valuable, really, several orders of magnitude than simply dumping several million lines of code onto someone. Also.. its funny that you mention RMS since he is so vehemently opposed to using the term open source :P