View Single Post
Old October 26th, 2012, 03:02 AM   #55 (permalink)
Speed Daemon
Join Date: Jul 2012
Posts: 2,033
Carrier: Sprint

Thanks: 541
Thanked 556 Times in 440 Posts

Originally Posted by EarlyMon View Post
Actually got tired of typing that on my phone.
I can imagine!

I did Mach development for two years. As first released (escaped is a better word) from academia, it had a number of significant bugs (in the endlessly nested manifest constants) and was incapable of real time instrument support, areas that I fixed and modified.

Yes. Mach is a microkernel.
You have my condolences. I heard a lot about Mach (and RISC) from my CS classmates when I was studying EE. At the time I didn't care, but when I got into serious OS support, I began to form my own opinions. I read AST's text on operating systems, and went on to play with Minix at the same time that I was using Linux.

The next time I heard about microkernels and Mach was when Apple was struggling to replace its ancient Mac OS. There were lots of rumors, including that Apple was moving to U of Utah's Mach4 There was so much empty hype back then.

I feel perfectly fine with the spelling cheats of unix, Unix and *nix.
I'm one of those people who has a pretty good idea how much of an impact that the UNIX® has made on everyone's lives, so when the Open Group says "this is how UNIX® is supposed to be used" I'm more than happy to oblige. IMHO UNIX® deserves a better fate than being reduced to a generic term like Xerox, and am willing to make the extra effort to keep the UNIX® brand alive. I'm also a very unhappy ex-Caldera shareholder.

And I assure you, OS X is based on the Mach MICROkernel. Before the public release (and in its first public incarnation), the Mach debug messages sang out loud and clear on any console bootup. I saw that quite often, with my own eyes. Pretty hard to mistake a Mach revision print statement.
I get your point, but I can't help but be reminded of when someone found some text about the Regents of the University of California in a MS-DOS file. Sure it proves something, but what?

I'll stipulate that OSX has a history steeped in CMU Mach, but wouldn't say that it's a working 100% microkernel-based system. Likewise, Windows NT was originally supposed to use a microkernel client/server architecture, but what made it to production was a compromise. And ever since, NT-based Windows has been using more and more monolithic technologies to improve performance. Pardon me for not saying with the insight that you have, but to me, removing the latency of state switching needed to pass every single message opens up a CPU to get useful work done.

As implemented for OS X, I believe that the hybrid is no longer a microkernel, as you say.
I hope I never said that! My understanding is that no OS design that's 100% faithful to the microkernel idea has succeeded in general purpose computing. There may be some embedded application that's ideal for it; I don't know.

My original point was that neither OSX or IOS came from AT&T UNIX®. But that's a tale for another day.
Speed Daemon is offline  
Reply With Quote