>FreeBSD is amazingly stable. Uptimes of a year are taken for granted.
People claiming uptimes of a year or more make me cringe. Don't they even CONSIDER applying the latest SECURITY PATCHES?
It depends... It does not have to be rebooted because of userland vulnerability. You pretty much may ignore local kernel exploits if there are no local users (except you;). And remote exploits happen not so often lately.
Aren't there certian optimizations and, in general, better coding for most 32 bit applications (on the lowest level of the code) because people have used it for so long? Couldn't it just be that we need to refine coding for 64 bit processors?
I believe gcc just isn't right tool to do the job.
64-bit optimizations in gcc are non-existent AFAIK.
2) Linux has a better, more advanced kernel. The original article author is wrong. The OS X kernel is not FreeBSD, but rather Mach 3.0 + 4.4BSD-Lite2. There are pieces imported from other BSDs (buffer cache, networking), but pieces like the block-IO, scheduler, and VM are still pretty outdated code.
Have you actually read article? It does not say OS X kernel is FreeBSD. It has a lengthy explanation of kernel structure and parts it consists of.
You are talking about outdated code - are you reading code and comparing it to Linux's or you are just believing it is outdated?
It's a little more than a FreeBSD jail, which looks like an almost-VM-but-not-quite. The UML is a complete VM that runs its own kernel (which can be different to the host, so it's good for experimenting), has its own memory allocation, own swap space etc. My physical server is split up into several UML instances - a little tiny one as a DNS server, a big one as a web server, a small one as a barebones webserver for static sites, another big one as a shell machine etc.
UML usefull for debugging, for some odd multipurpose servers, but not for production use. The main problem is memory. Each kernel allocates a lot of memory which could be unused on one kernel and overused on another. A lot of functionality dublicates in kernels..
In case of FreeBSD jails approach is more lightweight and security-wise is a-must on servers.
There is network stack virtualization project - something in between UML and jail -
http://www.tel.fer.hr/zec/vimage/
How can they include things like the correct errno.h for Linux into their closed source binaries without being in copyright violation? Remember that several of the Linux i386 values aren't POSIX compliant so SCO can't say they used the standards.
I'm just saying, I don't get the concept that merely making an O(1) scheduler is so incredibly difficult, and it's so utterly amazing that they were able to do so.
Well, they (afaik Ingo Molnar) created universal scheduler which should work properly on user workstation and on server, on uni- and multiprocessor systems with small latencies and without performane loss. It may be easy to create a model in your head, but proper implementation takes a lot of effort and keyword is 'proper'.
I wonder how different scheduler classes get scheduled. And how does scheduler differentiate load balancing for processes running on same core (SMT) and on different processors.
It would be nice to have more technical article published somewhere...
Hm, I just hit pause on the DVD and it starts right back up where it left off. You're not still watching that quaint old broadcast or cable stuff by any chance? Until TiVO can author its own shows, I don't think even it could make that steaming pile of reality shows and infomercials watchable.
Aww, common, some shows are bad and some are good - you just have to find something you like. By the way you may want to record pay-per-view movies or or movies which were not issued on DVD. And almost forgot - it's great for sports.
Which of course is rubbish. The BSD licence allows you to put a GPL copyright header on it, or even an MS EULA on it
Please differenciate coyright and license.
You can't change copyright on your own - only on agreement with author. You may change license but have to retain copyright notice.
Slashdot Mirror
Oh yea, remember that movie - Harley Davidson and the Marlboro Man? They had tracking device in dollar coin..
Is it true? I know about Theo being picky, but FreeBSD and NetBSD folks seem to be indifferent..
People claiming uptimes of a year or more make me cringe. Don't they even CONSIDER applying the latest SECURITY PATCHES?
It depends... It does not have to be rebooted because of userland vulnerability. You pretty much may ignore local kernel exploits if there are no local users (except you
Dude, look up numbers:
Company Market Cap (mil)
Oracle 71,617.94
PeopleSoft 7,873.70
SAP 54,914.42
Together Peoplesoft and SAP are smaller than Oracle...
Is it 110?
Do you have gcc for abacus ? ;)
By the way one more free software compiler is TENDRA - www.tendra.org (I have no idea about it's performance though).
I believe gcc just isn't right tool to do the job. 64-bit optimizations in gcc are non-existent AFAIK.
He should use Sun's compiler for benchmarks.
Opera works the same way.
Is it official Microsoft's position? ;))
Who's Geraldo Rivera here?
Gentoo Slackware
FreeBSD!
Yes, it was that bad.
That's not true. At that time M$ embedded IE into Windows and users just did not bother to download anything else.
You did not pay attention for what he said. He talked about NIC driver, not gigabit support per se. I was using gigabit with FreeBSD in 2000.
Have you actually read article? It does not say OS X kernel is FreeBSD. It has a lengthy explanation of kernel structure and parts it consists of.
You are talking about outdated code - are you reading code and comparing it to Linux's or you are just believing it is outdated?
Some people in Silicon Valley have it.
Why replace POP3 (and IMAP)?
Maybe he meant MAPI, which is used by M$ in Exchange and Outlook.
The smart thing to do would be to replace SMTP MTAs with something that does server-to-server authentication
Well, SMTP could go through SSL/TLS - using STARTTLS, or it could go in SSL tunnel (SMTPS).
UML usefull for debugging, for some odd multipurpose servers, but not for production use. The main problem is memory. Each kernel allocates a lot of memory which could be unused on one kernel and overused on another. A lot of functionality dublicates in kernels..
In case of FreeBSD jails approach is more lightweight and security-wise is a-must on servers.
There is network stack virtualization project - something in between UML and jail - http://www.tel.fer.hr/zec/vimage/
1997 ? Yeah, It's like you asked F-150 Ford owner if his truck could haul same amount of cargo as Toyota Corolla...
At that time Linux was very immature, especially in server area - almost everything was rewritten since then.
So if Tannenbaum got AT&T copyrighted files into Minix and Linux copied them - is it fine?
There is no copyright notice there.
Well, they (afaik Ingo Molnar) created universal scheduler which should work properly on user workstation and on server, on uni- and multiprocessor systems with small latencies and without performane loss. It may be easy to create a model in your head, but proper implementation takes a lot of effort and keyword is 'proper'.
However there are some stones unturned
I wonder how different scheduler classes get scheduled.
And how does scheduler differentiate load balancing for processes running on same core (SMT) and on different processors.
It would be nice to have more technical article published somewhere...
Aww, common, some shows are bad and some are good - you just have to find something you like. By the way you may want to record pay-per-view movies or or movies which were not issued on DVD. And almost forgot - it's great for sports.
well, if you miss a punchline..
Please differenciate coyright and license. You can't change copyright on your own - only on agreement with author. You may change license but have to retain copyright notice.