On 8/25/2014 3:27 AM, Stephan Aßmus wrote:
Hi Demetrious, I think you are blowing things out of proportion quite a bit and spreading FUD in the process. If you look a bit more at the details, you might come to a much more balanced understanding.
Hey, it's definitely not been FUD. Time will tell, though.
Which is exactly how some of the original BeOS drivers were written. But you'll notice that they still didn't use anyone else's kernel.The Haiku kernel has a lot of components and a lot of aspects. But take the network driver situation as an example. A FreeBSD source compatibility layer was developed, so that we can use all the FreeBSD network drivers. Some of the native drivers have been superseded by that, but looking even more closely, those had sometimes been ports of the FreeBSD driver in the first place.
How does this match up with R1? Before R1 is feature complete, how does this even matter? I seem to recall that this wasn't even a necessary feature in the beginning.What is the plan for getting 3D hardware acceleration into Haiku? It seems by writing more compatibility layers so that Linux drivers can be compiled for Haiku. Of course this is a huge undertaking and while multiple people have started it, no one has gotten results yet...
Time will tell. But I will tell you this -there's absolutely nothing wrong with sticking with the kernel that was worked on from the very beginning. I think a lot of you guys are being blinded by things that you don't yet have but aren't actually necessary for the feature completeness that you were originally going for. In this case, the interest seems to be in throwing away hard work that was carefully crafted from the very early days just to get features that don't bring the project any closer to having R1. But, do as you will. If you aren't going to stand up for the work that has already been done, then I guess I shouldn't do it either.Native graphics drivers are written by taking guidance and sometimes trying to sync parts of them with FreeBSD or Linux drivers. Same for some native audio drivers. We even have a port of Open Sound which sometimes has to be used to get sound on certain hardware. Take the Haiku multi-media add-ons. There were a few written, but most were actually wrappers to third party libraries. Where there were native plugins, they were inferior to FFmpeg. In the end, I poured more work into the FFmpeg plugin and basically replaced every other plugin that we originally had for Haiku by the one FFmpeg plugin. There is of course more to the kernel than just drivers. The VM layer and VFS would be most prominent. But ask any of our kernel developers and they will all agree that the speed and scalability of the Linux or FreeBSD kernel are in another league - and they would love for the Haiku kernel to play in the same! The code of the Haiku kernel might be much nicer and more approachable to new comers. But in terms of actual technical features, there isn’t actually a whole lot which makes the kernel special or even very different from the Linux or FreeBSD kernel. Now, in 2001, both Linux and FreeBSD still had the giant kernel lock. They didn’t have as much hardware support as they have now. They were probably even more ugly to dig into. Maybe they didn’t have some primitives that they have now which would have made implementing ports more difficult than now... Any lastly: If there was a true danger of people leaving the project if the Linux kernel would be seriously considered - why have most Haiku developers reacted at least interested in this thread? This would be just a parallel project for now, nothing is interrupted. Once it can be tried and tested, we can evaluate the question whether to absorb it. I don’t think the involved people would decide for something which is ultimately bad for the project or comes with too many compromises.
-- Regards, A. D. Sharpe, Sr.