That last comment has no basis in fact- syscalls are similar enough to the fact that binary compatibilty is not difficult to maintain.
That said, a unified ABI may not be too far away. Who knows?
-- -bugg
OB:Differences in *BSD
by
Anonymous Coward
·
· Score: 5
Since it gets asked everytime there's a *BSD article or release, might as well answer the differences between the flavours of BSD.
NetBSD is meant to run on as many platforms as possible. I've seen it run on old VAXen to Mac+'s, iMacs, x86, etc etc etc. Chances are if it's a halfway popular carchitecture, NetBSD will run in some way on it.
The install is a bit "different," but if you're looking for a *NIX for the old machine sitting in the corner try NetBSD.
OpenBSD is meant for security. It's the one with the line-by-line security code audit a few years back, and it shows. It's the only freenix I've ever installed, then not fiddled around with right afterwards so I wouldn't get rooted. I went and ate lunch instead. It's SMP is a bit lacking (but it's getting there;p), there are some complaints about the speed of it's file system, and it doesn't run on as many platforms as NetBSD.
I'm a bit biased since I cut my teeth on OpenBSD rather than Linux, so I can't speak if the install is easy or not anymore. But if you want security, get this.
FreeBSD is good for servers/workstations. It has good (hard to qualify) SMP, and runs on many good platforms. It can be made very secure easily, has an excellent ports tree for software, and is fastfastfast at networking.
The install is easy, the performance is good, and if you want to try out a *BSD but haven't before try FreeBSD.
It depends on your definitions of "operating system" and "distribution". Everyone has there own definitions. I like to draw lines between the operating system and the operating environment. Different distributions will have the same OS and nearly the same OE, but everything else can be different. In the case of Linux, the OS is the linux kernel plus necessary infrastructure (lilo, sysvinit, etc). The OE is mostly GNU.
In the case of the various BSDs, they all have different operating systems and operating environments. Comparing them to different Linux distributions misses the mark. Because they are completely different systems, they are further apart than the various Linux distros. But because they have kept close compatibility with each other, they can be closer together than some Linux distros are.
A real BSD "distribution" would be one that takes an existing BSD OS/OE and packages their own stuff on top. The only one that I know of is PicoBSD.
OT: I forsee the day when all these components are readily interchangable. Choose your OS (linux, bsd, hurd), choose your OE (GNU, BSD, mix), and choose your user software (shell, desktop, etc). Mix and match to your heart's content while running the installer.
-- A Government Is a Body of People, Usually Notably Ungoverned
Can the different types of BSD's be compared to the different distros of Linux? Or are they really different from each other? I've seen the list of so-called differences on http://www.bsd.com, but what are the REAL differences?
Whereas Linux is technically just a kernel, BSD is the kernel and the base system, including libc, basic utilities like cat, sed, grep, and so on (occasionally a GNU tool like tar is just too good to pass up, so it gets used instead). Unlike Linux, the three free flavors of BSD (Free, Net, Open) have different kernels and base systems, though they still borrow from each other. They can all run pretty much the same thing, and to a novice user running a desktop like kde or gnome, you really can't tell them apart from each other or even a Linux distribution (aside from the obvious differences in a desktop a vendor would supply).
NetBSD is something of a research platform for new ideas. New virtual memory and driver architectures have usually come out of NetBSD first. NetBSD is also amazingly portable, running on more architectures and devices than Linux has ever imagined. I don't know much about their organization structure, it seems to revolve around a self-organized core team. NetBSD's slogan is "Of course it runs NetBSD".
FreeBSD, founded at the same time of NetBSD (they were both forks of BSD/386) is aimed more at "real world" use. This isn't to imply that NetBSD isn't good for production use, but FreeBSD makes it their main focus, and thus eschews portability in favor of performance. It runs on only two platforms, alpha and ia32 (there's also a sparc port that's perpetually broken). FreeBSD uses a CVS tree to maintain the source, and the major developers can commit directly to the CVS tree. There is a core team of developers that can veto changes or make major changes affecting many systems; this team was just recently changed to an elected body, picked by all the committers. FreeBSD's slogan is "FreeBSD: The Power to Serve"
OpenBSD was a fork of NetBSD, but has grown into a full-fledged flavor in its own right. It's very unambiguous about its mission, and that's security. Every line of code in the distribution is audited for security holes, and encryption is ubiquitous, to the point of having features like encrypted swap partitions. OpenBSD's organization is most like Linux's in that it is run by a single person, Theo de Raadt, and a hand-picked team of developers, most of whom do the security audits.
All the BSD's are able to run Linux binaries at native speed. I personally run Linux Netscape on FreeBSD, because it supports Flash, and is more stable (which isnt saying much, I'm using IE to post this).
-- I've finally had it: until slashdot gets article moderation, I am not coming back.
Re:What happens when it gets popular?
by
moof1138
·
· Score: 3
I have NetBSD running on a Quadra 800, and an SE/30. There is a port of Linux to the hardware, but NetBSD is easier to install, more mature, and, for me, easier to admin. These machines make great little servers for low end stuff (print, DNS). I tried out the Linux m68k build about a year ago but I got so many kernel panics that I quickly determined that this was not an option.
I also use Darwin/OS X a lot, but again not because I dislike Linux, but because for what it is/does it is super.
BTW, personally I like how easy it is to admin a BSD system compared to a Linux box, I like the ports system more than rpm/apt. I find the portability of the OS handy as I have a lot of weird hardware, and being able to maintain some consistency is a good thing to keep administration simple. This is largely a matter of preference.
It is not fair to assume that NetBSDers ave a grudge against that more popular system for 'getting to big for its britches.' This is not about obscurantism. We like our lovely OS for what it is, and can do, not because that some other OS is too trendy.
I've spoken with people in the core of both the NetBSD and FreeBSD efforts (never had a chance to chat with OpenBSD folks). When I asked if there was a chance of merging the efforts and bringing the advantages of each into a single OS, I was sort of laughed at. Basically, the response was: there's way too much bad blood between us, and many of the core developers will spend weeks or months working on solutions to problems rather than even incorporate the idea that solved the problem in the other OS (e.g. the VM advances made in NetBSD vs. the VM work being done about a year ago in FreeBSD).
This is a sad state of affairs, and hurts BSD a lot.
You have the kernel. The kernel has a number of services, and a number of functions to do them.
You have the userland. The userland uses some libraries, and they, in turn, call the kernel to do some stuff.
Between them, this "call the kernel", is something named "syscalls". These are kernel "functions" that can be called from outside the kernel. They are the interface (in the _real_ CS sense of "interface") between kernel and userland.
Now... there is a large set of syscalls that is identical between most Unix, and there is a small set of syscalls that are different or unique to each OS. But these syscalls are _numbered_ differently.
So... what is the emulation, in the context of running Linux or FreeBSD binaries on OpenBSD (or NetBSD, since that's the focus of this story:)? Three things:
1) A numbering of syscalls identical to the OS being "emulated". There is *ZERO* speed difference here, because this is exactly what native applications do. The only difference between a native application and an imported binary is that different syscalls tables are selected depending on what is the binaries' type.
2) Some syscall "translation". This is the case where the native OS has a different order of arguments and/or more arguments or different flags values or some arguments/flags of the emulated syscall simply does not apply in the native OS. In this case, there simply is a call with the arguments reordered, and maybe some masking/translation of flags.
The cost of this is almost negligible, because the cost of *making* a syscall (crossing the userland/kernel border), in itself, is much greater already. Also, the different in speed of the *implementation* of said syscall will often far surpass, one way or another, this overhead.
3) Implementation of a feature non-existentent in the native OS. There is no overhead here, because there is no translation. The syscall is simply being implemented from the scratch. The implementation may be faster or slower than the emulated OS, because of the native OS' architecture. At the very least, Open Source means we can look at how the other OS implemented it.
And there you have. "Emulation", when used to speak of different hardware being emulated in software or different a different operating system being emulated, means slower. What is being done here is *not* emulation, but "ABI compatibility" just isn't catchy (or as easily understandable), so few people speak that way. What we do does _not_ add overhead.
As for databases and Java... Postgres and MySQL can be found in the ports (packages, on NetBSD) tree of all BSD. Java is more complicated, because it depends on a specific model of threads support. FreeBSD is, nowadays, mostly ok when it comes to Java (or so I'm told:). I do not know the state of Java on OpenBSD or NetBSD, but I'm afraid this specific application will not run emulated because it needs some very specific features support in the kernel. But don't take my word on it.
Actually, many BSDers *do* have a problem with restrictive licenses, and that's why they don't use Linux. Try suggesting on Open/Net/FreeBSD to replace a BSD-licensed util with a GPL-licensed one, for instance.
"But the BSD license let someone relicense it under a super-restrictive license," says someone.
Well, yeah. And, guess what?, we *don't use* that relicensed code, we stay with the BSD-licensed code.
What we don't have a problem with is other people using the code we created in whatever damned way they wish. That's the whole point of being non-restrictive.
-- (8-DCS)
Sad, not really [Re:Sad, really...]
by
jkh
·
· Score: 4
I don't know who ajs talked to, but I don't know of any "bad blood" between NetBSD and FreeBSD and haven't perceived any such thing for quite some time now. Yes, there were times over 7 or 8 years ago when inter-group relations were a little rocky but it very quickly became apparent to all of us that we were only shooting ourselves in the feet by engaging in sibling rivalry and those kinds of activities stopped.
Nowadays, there are a number of developers who sit astride both code bases and keep a close eye on development of import to both projects which can be shared. As someone already noted in this thread, the goals of both projects are also VERY different and a full "merger" just wouldn't make any sense at this point, nor would it even be particularly desirable.
I'm sure the people who do cancer research and the people who research cerebral palsy have a lot in common, for example, in that they're all doctors/health care professionals and share many of the same investigative methods and goals for improving the human condition. The fact remains that they've chosen to focus on different aspects of the very large set of problems facing humans today, however, and while they could probably derive some value in comparing notes from time to time, they're really going after different things.
So it is with FreeBSD and NetBSD. FreeBSD is focusing on providing a "product" to a wide variety of internet infrastructure and SOHO market folks and is driven by the demands of the market it's chosen to serve. NetBSD is focused more on OS research and providing a very high degree of OS portability. Both are very worthy and often complimentary goals, but they still require some fundamental differences in mind-set and the allocation of scarce (in both cases) volunteer resources.
-- - Jordan Hubbard
co-founder, the FreeBSD Project. Director, UNIX Technology. Apple Computer
Slashdot Mirror
That said, a unified ABI may not be too far away. Who knows?
-bugg
Since it gets asked everytime there's a *BSD article or release, might as well answer the differences between the flavours of BSD.
;p), there are some complaints about the speed of it's file system, and it doesn't run on as many platforms as NetBSD.
;)
NetBSD is meant to run on as many platforms as possible. I've seen it run on old VAXen to Mac+'s, iMacs, x86, etc etc etc. Chances are if it's a halfway popular carchitecture, NetBSD will run in some way on it.
The install is a bit "different," but if you're looking for a *NIX for the old machine sitting in the corner try NetBSD.
OpenBSD is meant for security. It's the one with the line-by-line security code audit a few years back, and it shows. It's the only freenix I've ever installed, then not fiddled around with right afterwards so I wouldn't get rooted. I went and ate lunch instead. It's SMP is a bit lacking (but it's getting there
I'm a bit biased since I cut my teeth on OpenBSD rather than Linux, so I can't speak if the install is easy or not anymore. But if you want security, get this.
FreeBSD is good for servers/workstations. It has good (hard to qualify) SMP, and runs on many good platforms. It can be made very secure easily, has an excellent ports tree for software, and is fastfastfast at networking.
The install is easy, the performance is good, and if you want to try out a *BSD but haven't before try FreeBSD.
Oh, and goatse.cx.
It depends on your definitions of "operating system" and "distribution". Everyone has there own definitions. I like to draw lines between the operating system and the operating environment. Different distributions will have the same OS and nearly the same OE, but everything else can be different. In the case of Linux, the OS is the linux kernel plus necessary infrastructure (lilo, sysvinit, etc). The OE is mostly GNU.
In the case of the various BSDs, they all have different operating systems and operating environments. Comparing them to different Linux distributions misses the mark. Because they are completely different systems, they are further apart than the various Linux distros. But because they have kept close compatibility with each other, they can be closer together than some Linux distros are.
A real BSD "distribution" would be one that takes an existing BSD OS/OE and packages their own stuff on top. The only one that I know of is PicoBSD.
OT: I forsee the day when all these components are readily interchangable. Choose your OS (linux, bsd, hurd), choose your OE (GNU, BSD, mix), and choose your user software (shell, desktop, etc). Mix and match to your heart's content while running the installer.
A Government Is a Body of People, Usually Notably Ungoverned
Can the different types of BSD's be compared to the different distros of Linux? Or are they really different from each other? I've seen the list of so-called differences on http://www.bsd.com, but what are the REAL differences?
Whereas Linux is technically just a kernel, BSD is the kernel and the base system, including libc, basic utilities like cat, sed, grep, and so on (occasionally a GNU tool like tar is just too good to pass up, so it gets used instead). Unlike Linux, the three free flavors of BSD (Free, Net, Open) have different kernels and base systems, though they still borrow from each other. They can all run pretty much the same thing, and to a novice user running a desktop like kde or gnome, you really can't tell them apart from each other or even a Linux distribution (aside from the obvious differences in a desktop a vendor would supply).
NetBSD is something of a research platform for new ideas. New virtual memory and driver architectures have usually come out of NetBSD first. NetBSD is also amazingly portable, running on more architectures and devices than Linux has ever imagined. I don't know much about their organization structure, it seems to revolve around a self-organized core team. NetBSD's slogan is "Of course it runs NetBSD".
FreeBSD, founded at the same time of NetBSD (they were both forks of BSD/386) is aimed more at "real world" use. This isn't to imply that NetBSD isn't good for production use, but FreeBSD makes it their main focus, and thus eschews portability in favor of performance. It runs on only two platforms, alpha and ia32 (there's also a sparc port that's perpetually broken). FreeBSD uses a CVS tree to maintain the source, and the major developers can commit directly to the CVS tree. There is a core team of developers that can veto changes or make major changes affecting many systems; this team was just recently changed to an elected body, picked by all the committers. FreeBSD's slogan is "FreeBSD: The Power to Serve"
OpenBSD was a fork of NetBSD, but has grown into a full-fledged flavor in its own right. It's very unambiguous about its mission, and that's security. Every line of code in the distribution is audited for security holes, and encryption is ubiquitous, to the point of having features like encrypted swap partitions. OpenBSD's organization is most like Linux's in that it is run by a single person, Theo de Raadt, and a hand-picked team of developers, most of whom do the security audits.
All the BSD's are able to run Linux binaries at native speed. I personally run Linux Netscape on FreeBSD, because it supports Flash, and is more stable (which isnt saying much, I'm using IE to post this).
I've finally had it: until slashdot gets article moderation, I am not coming back.
I have NetBSD running on a Quadra 800, and an SE/30. There is a port of Linux to the hardware, but NetBSD is easier to install, more mature, and, for me, easier to admin. These machines make great little servers for low end stuff (print, DNS). I tried out the Linux m68k build about a year ago but I got so many kernel panics that I quickly determined that this was not an option.
I also use Darwin/OS X a lot, but again not because I dislike Linux, but because for what it is/does it is super.
BTW, personally I like how easy it is to admin a BSD system compared to a Linux box, I like the ports system more than rpm/apt. I find the portability of the OS handy as I have a lot of weird hardware, and being able to maintain some consistency is a good thing to keep administration simple. This is largely a matter of preference.
It is not fair to assume that NetBSDers ave a grudge against that more popular system for 'getting to big for its britches.' This is not about obscurantism. We like our lovely OS for what it is, and can do, not because that some other OS is too trendy.
Hyperbole is the worst thing ever.
I've spoken with people in the core of both the NetBSD and FreeBSD efforts (never had a chance to chat with OpenBSD folks). When I asked if there was a chance of merging the efforts and bringing the advantages of each into a single OS, I was sort of laughed at. Basically, the response was: there's way too much bad blood between us, and many of the core developers will spend weeks or months working on solutions to problems rather than even incorporate the idea that solved the problem in the other OS (e.g. the VM advances made in NetBSD vs. the VM work being done about a year ago in FreeBSD).
This is a sad state of affairs, and hurts BSD a lot.
Let me explain here how does "emulation" work.
:)? Three things:
:). I do not know the state of Java on OpenBSD or NetBSD, but I'm afraid this specific application will not run emulated because it needs some very specific features support in the kernel. But don't take my word on it.
You have the kernel. The kernel has a number of services, and a number of functions to do them.
You have the userland. The userland uses some libraries, and they, in turn, call the kernel to do some stuff.
Between them, this "call the kernel", is something named "syscalls". These are kernel "functions" that can be called from outside the kernel. They are the interface (in the _real_ CS sense of "interface") between kernel and userland.
Now... there is a large set of syscalls that is identical between most Unix, and there is a small set of syscalls that are different or unique to each OS. But these syscalls are _numbered_ differently.
So... what is the emulation, in the context of running Linux or FreeBSD binaries on OpenBSD (or NetBSD, since that's the focus of this story
1) A numbering of syscalls identical to the OS being "emulated". There is *ZERO* speed difference here, because this is exactly what native applications do. The only difference between a native application and an imported binary is that different syscalls tables are selected depending on what is the binaries' type.
2) Some syscall "translation". This is the case where the native OS has a different order of arguments and/or more arguments or different flags values or some arguments/flags of the emulated syscall simply does not apply in the native OS. In this case, there simply is a call with the arguments reordered, and maybe some masking/translation of flags.
The cost of this is almost negligible, because the cost of *making* a syscall (crossing the userland/kernel border), in itself, is much greater already. Also, the different in speed of the *implementation* of said syscall will often far surpass, one way or another, this overhead.
3) Implementation of a feature non-existentent in the native OS. There is no overhead here, because there is no translation. The syscall is simply being implemented from the scratch. The implementation may be faster or slower than the emulated OS, because of the native OS' architecture. At the very least, Open Source means we can look at how the other OS implemented it.
And there you have. "Emulation", when used to speak of different hardware being emulated in software or different a different operating system being emulated, means slower. What is being done here is *not* emulation, but "ABI compatibility" just isn't catchy (or as easily understandable), so few people speak that way. What we do does _not_ add overhead.
As for databases and Java... Postgres and MySQL can be found in the ports (packages, on NetBSD) tree of all BSD. Java is more complicated, because it depends on a specific model of threads support. FreeBSD is, nowadays, mostly ok when it comes to Java (or so I'm told
(8-DCS)
Actually, many BSDers *do* have a problem with restrictive licenses, and that's why they don't use Linux. Try suggesting on Open/Net/FreeBSD to replace a BSD-licensed util with a GPL-licensed one, for instance.
"But the BSD license let someone relicense it under a super-restrictive license," says someone.
Well, yeah. And, guess what?, we *don't use* that relicensed code, we stay with the BSD-licensed code.
What we don't have a problem with is other people using the code we created in whatever damned way they wish. That's the whole point of being non-restrictive.
(8-DCS)
I don't know who ajs talked to, but I don't know of any "bad blood" between NetBSD and FreeBSD and haven't perceived any such thing for quite some time now. Yes, there were times over 7 or 8 years ago when inter-group relations were a little rocky but it very quickly became apparent to all of us that we were only shooting ourselves in the feet by engaging in sibling rivalry and those kinds of activities stopped.
Nowadays, there are a number of developers who sit astride both code bases and keep a close eye on development of import to both projects which can be shared. As someone already noted in this thread, the goals of both projects are also VERY different and a full "merger" just wouldn't make any sense at this point, nor would it even be particularly desirable.
I'm sure the people who do cancer research and the people who research cerebral palsy have a lot in common, for example, in that they're all doctors/health care professionals and share many of the same investigative methods and goals for improving the human condition. The fact remains that they've chosen to focus on different aspects of the very large set of problems facing humans today, however, and while they could probably derive some value in comparing notes from time to time, they're really going after different things.
So it is with FreeBSD and NetBSD. FreeBSD is focusing on providing a "product" to a wide variety of internet infrastructure and SOHO market folks and is driven by the demands of the market it's chosen to serve. NetBSD is focused more on OS research and providing a very high degree of OS portability. Both are very worthy and often complimentary goals, but they still require some fundamental differences in mind-set and the allocation of scarce (in both cases) volunteer resources.
- Jordan Hubbard co-founder, the FreeBSD Project. Director, UNIX Technology. Apple Computer