Posted on 10/04/2010 7:59:37 AM PDT by ShadowAce
So you’re saying that MS is to blame because it delegates
driver development to the hw vendors? Well, they (the vendors) do WHCL (Windows Hardware Compatibility List) testing on their products to assure MS that they meet
a certain standard of usability. They can then put the
MS logo in the box. The rest can still program to the
DDK (a public interface) and MS takes no position on that code’s reliability. If I am developing a new HW product
I can dive in and try to make it work without having to
beg MS to give me an interface. What does kind of free kernel access does Apple provide? Do they provide a free DDK? Or do I need to sign a NDA, get layers of approvals and pay $$ for the information?
Microsoft didn't delegate in the first place. What MS did is allow those device developers high-level access to the kernel where before they were put in userland where they couldn't do much damage. Once Microsoft decided to do that, their business model is what killed them, although it is now less of a problem because all drivers must be signed (and supposedly tested). Apple also gives privileged access to drivers, but Apple controls the hardware and drivers, making it less of a problem.
What does kind of free kernel access does Apple provide?
In my perfect world, NOBODY would give any drivers kernel access. Take RIM's upcoming tablet that will run QNX, which is an extremely stable microkernel system. Device drivers (including video and network) run in userland, and can be killed and restarted if they become unstable or crash without affecting the rest of the system.
-In my perfect world, NOBODY would give any drivers kernel access.-
That would certainly increase reliability. But I wonder
if such a model would be appropriate for primary storage and video. Some sort of DMA access would still be needed
for these devices to function at acceptable speed.
Andrew, is that you? <grin>
It's happening all the time, and monolithic/hybrid kernels have to timeslice that anyway. Integrity and QNX are examples in use today in the commercial market. They are also the only kind of operating systems that may be rated EAL7 one day. Windows will never get higher than its current EAL4, and neither will most operating systems, but Integrity is at EAL6+. The seL4 microkernel has even formally verified, and the only operating systems to be evaluated at EAL7 are microkernel-based (it's pretty much impossible for a monolithic kernel to get EAL7).
I think that one was part of Apple's lawsuit. Expect that Apples window says You computer has become unstable and needs to see an expensive therapist...
Okay, I admit that much of my early OS design knowledge I learned from him. Literally, he wrote my OS design college textbook. He wrote my networking textbook too. But most of my views in this area come from me being an anal-retentive neat freak when it comes to computing. Torvalds didn't want to tackle the performance issues of microkernels, thus advancing computer science, so he just went monolithic with something anybody could have built. And like any good professor, Tanenbaum called him on it.
There was a massive flood of Microsoft employees that headed to Google. I wonder if any of that was a result.
i can. Of course it had nothing to do with SuSE Linux. The fault was a defective stick of RAM.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.