On Mon, Aug 18, 2014 at 5:19 PM, Stephan Aßmus <superstippi@xxxxxx> wrote: > > You have the benefit of hindsight. First of all, there were indeed some > technical reasons -- ports, inode ids in entry_ref, ... there was more, but > I am no kernel guy. And it was believed that having to maintain a patch set > on top of the Linux kernel would be too inconvenient. There was the > argument of being in full control of everything top to bottom. And the > Linux source code itself is probably in a much better state today than it > was in 2001. It is probably still ugly and cryptic in many places compared > to Haiku. Though I don't really have a clue. > > Anyway. If you keep going, you will probably stumble across some of the > reasons why Haiku went with its own kernel. Maybe all of it can be worked > around if binary compatibility is not your goal. (Personally, I think it > doesn't have to be.) Who knows, maybe the Haiku project is prepared to > switch kernels once you are further along? > Whatever the reasons were back then, I've come far enough to reach this important conclusion: No patch needs to be maintained, and you can easily draw a "don't cross this line" at the kit->kernel/composer transition layer (not without tricks, but they're completely self-contained in a kernel module). Hence, you can still use the C++ toolchain with EH runtime support and all that. There are no "upstream issues" since there's nothing to be patched. Unsurprisingly, the same would be true for, say, a FreeBSD kernel. Anyway, the proof is in the pudding, so let's go to work ;) Sia.