“There’s been a lot of talk about the Mac OS X kernel in the past few months. When Avie Tevanian left Apple in March, there was much speculation that Apple had plans to abandon the Mach microkernel that underpins Mac OS X, and that Avie helped to develop many years ago. There were two lines of reasoning to support this, both pretty silly.” Update: Follow-up has been posted. Thanks to eMagius pointing that out.
or is this article a bit silly – a whole page of this person thinking that there is something wrong – but not really any facts of what is going on .
This is IMO not worth reading.
A whole page of this guys opinion without any reaons for it .
We have many kinds of different stuff .. lots of stuff in this “article” .
It lacks content .
😉
EDIT : THX for follow-up – it still lacks content .
He portrays himself as someone with special insider knowledge predicting abilities but is still just an attention seeker IMO .
& Why does the open source kernel version need to be “pretty” ?
I could have written these BLOG entires but refrain from it because I know that I do not have enough knowledge to have an informed opinion to spread via blogs .
Edited 2006-06-19 19:17
John Siracusa is famous in the industry for his extraordinarily insightful previews and reviews of Mac OS X during the entire course of its development. His review of Tiger (http://arstechnica.com/reviews/os/macosx-10.4.ars) if printed, would take 112 pages.
He never claims to have any insider knowledge: he just examines Apple’s past behaviour, considers their current position and the amount of press at the moment, and from that extrapolates a possible outcome.
Pretty, in this case refers to marketing opportunities.
It is ironic that you say you refrain from blogging on issues of which you are ignorant, but then wrote an entire comment criticising an author of whom you knew nothing.
Personally I don’t think it’s likely, as I don’t think low-level details are something that Apple considers very much. None of the existing micro-kernels are drop-in replacements: it would take an entire re-write to do that. He’s right that stabilising the API via K-Exts is the first step, but I think changing the kernel would be such a huge job that we wouldn’t see it till Mac OS XI.
Edited 2006-06-20 10:08
http://arstechnica.com/staff/fatbits.ars/2006/6/19/4376
While I agree the article lacks depth, I also agree that there really isn’t anything about the OS X kernel that is really all that bad to warrant blowing it out in favor of the Linux kernel. Using Linux would require way too much modification for it to be worth it, not to mention having to write all new drivers for the OS. The Linux kernel is just not built for home desktop use, especially since drivers can be severely broken with every revision.
The linux kernel is fine for home desktop use. That’s a fairly broad statement based on the fact that there are tons of distros out there without proper support. Apple would not let you download and compile the latest kernel. I have no doubt they would limit the ability for the user to break the system so badly just for doing “updates”.
Obviously Apple wouldn’t be so brazen to follow the mainline Linux kernel tree. There are companies out there (RedHat, for example) that intentionally use older, well-proven kernels to ensure binary compatibility with existing drivers. RedHat Enterprise Linux doesn’t see kernel updates every time Andrew Morton does a kernel release. Apple would be no different.
With that said, i don’t see them moving to Linux for the simple fact that I think some developers are still reeling over the migration to Intel. They don’t want to have to rewrite their drivers yet again.
If anything apple would move on a bsd kernel, since they could re-licence it as they saw fit, also they employ a number of bsd developers full time so it would also be a logical fit with whom they keep in-house.
But I have to agree with what others have said, a kernel switch doesn’t seem the most likely move. Especially, as EmmEff points out with the intel switch, then again apple moves mysterious ways.
While I agree the article lacks depth, I also agree that there really isn’t anything about the OS X kernel that is really all that bad to warrant blowing it out in favor of the Linux kernel. Using Linux would require way too much modification for it to be worth it, not to mention having to write all new drivers for the OS. The Linux kernel is just not built for home desktop use, especially since drivers can be severely broken with every revision.
Screw the so-called ‘suitability’ – the licence is the blocking factor, before one even gets to the technical arguments.
Personally, the whole debate is nothing more than a sad parlour game – the fact is, yes, there are bottle necks in reference to performance, but isn’t as though one needs to throw the baby out with the bathwater to squeeze some extra performance out of it.
Edited 2006-06-20 05:06
The OS X Kernel hasn’t been release under de Apple Public License ( APPL ) due to the Transitive License Agreement for Rosetta. That is greatly possible ( I think this is the only ‘speculation’ that is possibly really true ).
For the other pieces of speculation… IMHO I think it’s really really premature to make such claims… I would welcome a Kernel Upgrade if it would mean boost the performance of the whole OS! But that is only wishful thinking…
My bet is that OS X Kernel will likely remain the same without much changes (Apart from a couple of enhancements/fixes)
My 2 cents.
…is NOT going to swap out the kernel. Even if they got the notion to do so – it would take years to accomplish that and it would be one of those famous Apple “secrets” that hit you about a month before they released it, suspiciously leaked to generate curiosity.
Well, they already switched to x86. I don’t think that is a smaller feat than changing the kernel. And maybe they have been working on another kernel for years.
That is not 100% accurate. They HAVE been building against intel for years BEFORE they switched.
Please, please, please no FreeBSD kernel.
L4 is an interesting thought, but cobbling Mach on top of it seems like YAFA (Yet Another Fine Abstraction), and I’d imagine that you wouldn’t gain any significant performance benefits. L4’s IPC is quite a bit different than Mach’s, by design. And Mach isn’t something you could just nicely emulate on top of it using native calls. If anything, you’d probably get better performance by porting NetBSD w/ Mach emulation. But adding another abstraction layer would also allow things like parallelism and the ability to run a second OS simultaneously ala Xen (hello Vista for L4).
The main problem with taking netbsd, the recently releases of which have been showing some outstanding performance numbers, is that its SMP support is behind that of freebsd or linux. They are operating under a central giant lock, similar to freebsd 4.x.
Given the increasing move to dual, and more, core cpus either freebsd or linux are in a far better position to deal performance over the additional cores.
I understand that, however….
I’ve been completely unimpressed with FreeBSD later than 4.x; it still refuses to work on hardware where 4.x ran flawlessly, and with the benchmarks I’ve seen, the UP performance needs lots of help. Apple isn’t going to choose Linux because of GPL concerns.
I’d have to think about it a bit more….the implications of throwing something in on top of L4. I’m thinking that the native SMP performance wouldn’t matter as much, because that would be left to L4. (L4-Linux works in SMP-mode with Pistaschio, but not with earlier versions of L4)
DragonFly also hasn’t gotten a mention. I’d imagine serializing their LWKT model to L4 would be more straightforward than dealing with FreeBSD.
Caveat, I haven’t heard anything about the Apple kernel in the last few months, or years.
Apple has to be doing some ongoing kernel hacking to support new features & secutity, or has there been no new features. Rumor has it Firewire 800 won’t be part of the Intel Apple products, that seeems to be a step backards, and hFW 800 is available in the x86 PC world.
Uhmm.. Based on your comment I assume you’re not aware the 17″ MBP’s have FW800. So your rumour is an ancient one, at best.
Mach has demonstrated portability and flexibility. Why break a good thing?
Neal Saferstein
Matthew Dillon talked about DF at Apple campus:
http://leaf.dragonflybsd.org/mailarchive/kernel/2005-12/msg00024.ht…
Here is the presentation:
http://www.dragonflybsd.org/docs/LISA200512/