On Fri, Aug 22, 2014 at 5:42 PM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote: > BTW, if you want an example, please have a look at the Intel AGP bus > driver to see how complex you can solve simple things to the point that > it's very hard to understand what's going on. > You have to do better than pick a single driver (that particular one is mostly just pci register read/writes anyway). The vast majority of Linux drivers are very very good, and some of the apparent complexity comes from stuff lacking in Haiku, such as proper sleep/wake behavior, power management, firmware workarounds, black/white listing of chipset variations ... the list goes on. So while these mundane details admittedly uglyfi the code, in a real kernel there's no way around it. > Don't get me wrong; it's probably the best open source kernel available, > but that doesn't mean it's perfect, or doesn't have issues in areas I do > care about. > > Hmm, so what's the rational thing to do? > I completely understand that "fun" is the main motivator, and if "ground >> up" is a requirement for the fun part, then so be it. It just means >> we'll never see Haiku released, which from a prospect end-user >> perspective is rather sad. >> > > You completely disregard how far we've come already; we actually did build > it from the ground up. That's certainly an explanation why it took so long > to get where we are now, but it's not an excuse to never release Haiku. I'm > afraid you'll have to come up with some new arguments for that claim. > Let time show I guess, but the point is this: if you've walked 450 miles and the target is moving away from you at the same speed... you've not walked very far, now have you? That's the reality of Haiku's approach and why I think Phipps believed we were "almost there" 7 years ago. Sia. P.S. There's a lot of amazing work done in Haiku, my points are about the kernel/driver situation (and package manager, arrrgh)