I seriously doubt they're interested in an OS as we know it.
More likely they want a framework for network applications to be available on the client side with more fidelity than you can get on a web browser. That's exactly the sort of thing Google would do.
Building on Firefox to do this is a pretty easy way to get where they want. Javascript is a very clumsy way to put code on the client side, while Java applets tend to push too much of the logic onto the client side and the sandbox they live in disconnects them from the browser too much.
My guess is that they want a middle ground -- much more powerful than Javascript, but without the restrictions of Java applets. Simply customizing Firefox is a very quick way to get there. They can create any scripting language they want with any API they want, and it will be available to every platform that runs Firefox instantly. That's what I call deploying software.
Think about that. You know those 1U Google search appliances they sell? Imagine one of those hosting your application, with much more advanced client side scripting than any browser can do now, an API on the server to make it easy and convenient for developers to make it go, and working perfectly on every computer in the company no matter what OS, version, patch level, etc.
That's exactly what the guy was talking about. Deploying software.
People that are good at designing robust, orthogonal, extensible APIs are few and far between. If that's what Google wants to do with him, they can get their money's worth and more without ever touching any Microsoft IP.
As a programmer, I find the concept of programmers being infallible hilarious. There's a reason we call it "cvs blame" at work instead of "cvs annotate".
I find I get a lot done on OpenBSD because of the simplicity. Also, for C code I find that the OpenBSD libraries do the fewest weird things, so it's easier to write something that needs to run on Linux, Solaris, AIX, etc.
Wherever I'm working, I've usually got browser windows for docs, SSH sessions to my OpenBSD box, and emacs windows tunneled from my OpenBSD box. I prefer my Linux box for these purposes because it has two big monitors instead of one small one like my iBook, and it's a little awkward having half of your windows coming from X11.app on a Mac, but the difference isn't that big.
This is just me doing C development... obviously for someone with different habits and different needs the results could be different.
"True, but a plausible argument can be made that BSD style freedom permits a great derived work to appear only in closed source code that can achieve success in broad binary distributions. The improved source code base is unavailable."
There are two classes of patch that mitigate this. The majority of patches fall into this category IMO.
First, there's generic bugfixes. These are typically submitted back to the community even with BSD licenses because companies don't feel like maintaining a private patch set. A company I used to work for is responsible for lots of updates to Python due to this. Sure we could have a better Python than the official distribution, but we wouldn't get paid for that so it was a waste of our time to keep the fix to ourselves.
Second, there's private customizations. The GPL explicilty allows these as well. These typically aren't useful to the community at large, because they're application specific, and patches related to them wouldn't be accepted. This happend with IBM patches to GCC for PowerPC970 optimizations that were rejected. The only difference is that the BSDl allows binaries to be distributed without source in this case.
The stuff that you really want to get back the community is the large scale optimizations, but these are typically written by a core of people that don't have anything to do with a company, except for occasional funding.
There are only a very few projects that are big enough to get changes that benefit the community by private companies. Apache, the Linux kernel, that sort of thing. In these cases the GPL is indeed preferable.
In other cases, it seems as though a small group of talented developers gets more done almost to the exclusion of community involvement or license. DragonFlyBSD and OpenBSD have done amazing things with small communities, as have the GNU people with HURD. BeOS was an amazing accomplishment in an entirely proprietary environment.
I'm glad the GPL is available, but there's room in the world for other licenses.
It doesn't run on the really big machines, but then neither does MacOS. So if I'm wrong about enterprise class hardware, then the person I was responding to is also wrong about enterprise class because the best XServe doesn't compare to the machines I've linked to.
"FreeBSD doesn't even understand NUMA, which is basically indispensable for Opteron, POWER, or any serious Intel IA64 or x86 platforms."
FreeBSD does understand NUMA with the ULE scheduler in 5.x.
I'd be shocked if MacOS X understood NUMA though, as Apple has never sold a NUMA box.
"Gentoo's portage doesn't even come close to the flexibility and reliability of ports."
Portage's flexibility is considerably better.
However I agree on the other point. Portage's reliability is much, much... much worse.
Re:Does FreeBSD really need to prove itself?
on
The Case for FreeBSD
·
· Score: 3, Insightful
"so of what consequence are NetBSD's criticisms?"
Just because NetBSD has fewer users doesn't mean its criticisms are without consequence. After all, by that logic FreeBSD's criticisms of Linux would also be without consequence.
"The freedom of BSD has the danger of making you a prisoner of its distributed derivatives."
How? If you don't like the version the company you're dealing with (Sun, Apple) is shipping, you can always get the official software from openssh.org.
"GPL code belongs to you for the asking. That is also why GPL will eventually out-evolve all other software."
No. What has become obvious is that the community of developers is what drives the evolution of a system. Either can stagnate, either can advance quickly.
That is exactly what IBM doesn't do. They'll work with you on any platform you can come up with because they're primarily a consulting company.
They may have sold their PC business because it wasn't profitable, but they'll still hook you up with a PC if you want. Or a Mac. Probably even a Solaris box.
I think there's a good chance Apple will license OS X to IBM for high end workstations, despite their policy and everything that's been said. PowerMacs don't have the features necessary to be used for high end workstation tasks in a lot of places (For example, lack of ECC memory. Yes, I know you've never had problems with Apple memory, but that doesn't make ECC optional for mission critical jobs.).
Apple's not going to make custom PowerMacs for niche markets, but IBM could and OS X would be a sweet edition for that.
A workstation edition of MacOS could easily command a much higher price, so Apple would keep their margin and expand into markets they can't touch right now.
"If OS X is the only real desktop alternative, nothing is stopping people from buying Macs you know."
For every "big mistake" that was going to kill Apple and didn't, there's a killer app that was going to take over the world and didn't. About the only world they've taken over is the portable music player market, and that's platform agnostic (even if some people switch as a result).
Hell that's what I thought the mini would be during that orgy of speculation in the week or two before the keynote. A small box that can leverage a large box for media purposes.
I also speculated that Apple might offer a downloadable movie service. Even if you have to start the download 1/2 an hour early to let it buffer, that's still better than going to Blockbuster.
The bandwidth costs would be hefty for Apple, and having some kind of p2p thing doesn't seem like their style even if they made it seamless and DRMed.
I dunno.
With Apple you can guess price points but you can't guess the crazy shit they will pull out of their asses, and you also can't guess the obvious Good Things that they will pass up.
Also, despite Oppenheimer's claims they won't be licensing MacOS X to OEMs I think there's a very good chance they'll license it to IBM for high end workstations that won't compete with PowerMacs.
-The prices Dell gets for Intel processors and chipsets and other parts make AMD look expensive. -Dealing with greater variety of processors and other parts increases overhead for manufacturing and support. -Prejudice in the industry left over from earlier generations of AMD chips that weren't as solid. -AMD chips are performing very well everywhere right now, but the difference is big enough for big customers to start switching mostly on big multi-processor systems, which is a tiny minority of all the computers out there (though the margins are nice).
Using rand(3) would be part of the problem. It's known for having low randomness for the low order bits, at least older and less complicated implementations are. That is very likely the cause of the problem.
I wrote a little program to shuffle lines from stdin with drand48 seeded from/dev/urandom... the resulting playlists still seemed to have clumping. I dunno.
I really don't care enough to conduct anything approaching the kind of test it would take to verify that kind of thing.
It is indeed treated as a standard USB mass storage device. There's a setting for how much space to reserve for user files. It's FAT32. I don't have any Windows machines but I can mount it on my Linux machine just fine. Can't seem to get it to charge on Linux though, I think it needs software on the host machine to charge. Hoepfully the gtkpod people will update soon.
Battery is rated at 12 hours but I seem to be able to get more, and apparently using AAC 128 kbps extends it. I have iTunes transcode my high bitrate MP3 files to AAC for this reason (also to increase the number of songs). There's generation loss, but the second generation files in the shuffle are never copied anywhere else so it's okay.
Well, it's really annoying when certain songs come up over and over while others go unplayed, especially if you like the ones that are missed.
I had a bad MP3 CD player (not an Apple product, as they've never made such a device) that would repeat some songs with a probability of maybe 1 in 20, and there were other songs that went unplayed unless I specifically picked them. With under a hundred songs and me going months between burning new disks, the probability of that is very low with a good pseudo-random function.
My iPod seems to favor some bands, but that could be just me. It's certainly not as bad as that piece of crap CP player.
The obvious thing to do is probably to gather a statistically significant number of trials and see what the distribution is then. Any bias would become apparent.
"Computers can't generate true random numbers (ok, at least I don't know of any current methods) but only pseudo random numbers. There's a precise mathematical description that gets you from one number to the next."
No shit.
But they can gather entropy from lots of sources. For example, the timing of user interactions.
It can continue to generate electricity at night because heat energy is stored at the bottom, and as the environment cools at night (as it does fairly quickly in cloudless environments like that) the stored heat can allow the cooler air to rise.
Slashdot Mirror
I seriously doubt they're interested in an OS as we know it.
More likely they want a framework for network applications to be available on the client side with more fidelity than you can get on a web browser. That's exactly the sort of thing Google would do.
Building on Firefox to do this is a pretty easy way to get where they want. Javascript is a very clumsy way to put code on the client side, while Java applets tend to push too much of the logic onto the client side and the sandbox they live in disconnects them from the browser too much.
My guess is that they want a middle ground -- much more powerful than Javascript, but without the restrictions of Java applets. Simply customizing Firefox is a very quick way to get there. They can create any scripting language they want with any API they want, and it will be available to every platform that runs Firefox instantly. That's what I call deploying software.
Think about that. You know those 1U Google search appliances they sell? Imagine one of those hosting your application, with much more advanced client side scripting than any browser can do now, an API on the server to make it easy and convenient for developers to make it go, and working perfectly on every computer in the company no matter what OS, version, patch level, etc.
That's exactly what the guy was talking about. Deploying software.
People that are good at designing robust, orthogonal, extensible APIs are few and far between. If that's what Google wants to do with him, they can get their money's worth and more without ever touching any Microsoft IP.
Which BSD? I tried it on my OpenBSD 3.6 box. Or is this that new OpenCVS project?
As a programmer, I find the concept of programmers being infallible hilarious. There's a reason we call it "cvs blame" at work instead of "cvs annotate".
I find I get a lot done on OpenBSD because of the simplicity. Also, for C code I find that the OpenBSD libraries do the fewest weird things, so it's easier to write something that needs to run on Linux, Solaris, AIX, etc.
Wherever I'm working, I've usually got browser windows for docs, SSH sessions to my OpenBSD box, and emacs windows tunneled from my OpenBSD box. I prefer my Linux box for these purposes because it has two big monitors instead of one small one like my iBook, and it's a little awkward having half of your windows coming from X11.app on a Mac, but the difference isn't that big.
This is just me doing C development... obviously for someone with different habits and different needs the results could be different.
"True, but a plausible argument can be made that BSD style freedom permits a great derived work to appear only in closed source code that can achieve success in broad binary distributions. The improved source code base is unavailable."
There are two classes of patch that mitigate this. The majority of patches fall into this category IMO.
First, there's generic bugfixes. These are typically submitted back to the community even with BSD licenses because companies don't feel like maintaining a private patch set. A company I used to work for is responsible for lots of updates to Python due to this. Sure we could have a better Python than the official distribution, but we wouldn't get paid for that so it was a waste of our time to keep the fix to ourselves.
Second, there's private customizations. The GPL explicilty allows these as well. These typically aren't useful to the community at large, because they're application specific, and patches related to them wouldn't be accepted. This happend with IBM patches to GCC for PowerPC970 optimizations that were rejected. The only difference is that the BSDl allows binaries to be distributed without source in this case.
The stuff that you really want to get back the community is the large scale optimizations, but these are typically written by a core of people that don't have anything to do with a company, except for occasional funding.
There are only a very few projects that are big enough to get changes that benefit the community by private companies. Apache, the Linux kernel, that sort of thing. In these cases the GPL is indeed preferable.
In other cases, it seems as though a small group of talented developers gets more done almost to the exclusion of community involvement or license. DragonFlyBSD and OpenBSD have done amazing things with small communities, as have the GNU people with HURD. BeOS was an amazing accomplishment in an entirely proprietary environment.
I'm glad the GPL is available, but there's room in the world for other licenses.
"FreeBSD runs none of them, so what enterprise class hardware *does* it run on?"
HP Proliant DL585
Sun Fire V40z
It doesn't run on the really big machines, but then neither does MacOS. So if I'm wrong about enterprise class hardware, then the person I was responding to is also wrong about enterprise class because the best XServe doesn't compare to the machines I've linked to.
"FreeBSD doesn't even understand NUMA, which is basically indispensable for Opteron, POWER, or any serious Intel IA64 or x86 platforms."
FreeBSD does understand NUMA with the ULE scheduler in 5.x.
I'd be shocked if MacOS X understood NUMA though, as Apple has never sold a NUMA box.
It's better at handling more diverse dependencies and upgrading existing software.
It's also more up to date, but that's a very mixed blessing, because they don't do proper regression testing and stuff breaks a lot as a result.
" Oh, really. I run OS X Server, with Darwin Ports, with Open Directory, etc., etc., etc.."
Darwin Ports and Fink are 3rd party addons, not supported by Apple.
"It blows FreeBSD away in ease of administration"
Ease of administration comes down to opinion.
"comes shipped on enterprise-class hardware"
FreeBSD can run on enterprise class hardware if you feel like buying enterprise class hardware.
"and runs the same software under the hood."
Except for the kernel, which is different, and the cool features like jails and PF.
"Why FreeBSD? Why go half-way? Why not switch to Mac OS X? It has everything that FreeBSD has"
Except PF, jails, ports, etc.
MacOS is an excellent desktop OS, but it can't touch FreeBSD as a server. Even the server edition is behind FreeBSD.
"Gentoo's portage doesn't even come close to the flexibility and reliability of ports."
Portage's flexibility is considerably better.
However I agree on the other point. Portage's reliability is much, much... much worse.
"so of what consequence are NetBSD's criticisms?"
Just because NetBSD has fewer users doesn't mean its criticisms are without consequence. After all, by that logic FreeBSD's criticisms of Linux would also be without consequence.
"The freedom of BSD has the danger of making you a prisoner of its distributed derivatives."
How? If you don't like the version the company you're dealing with (Sun, Apple) is shipping, you can always get the official software from openssh.org.
"GPL code belongs to you for the asking. That is also why GPL will eventually out-evolve all other software."
No. What has become obvious is that the community of developers is what drives the evolution of a system. Either can stagnate, either can advance quickly.
That is exactly what IBM doesn't do. They'll work with you on any platform you can come up with because they're primarily a consulting company.
They may have sold their PC business because it wasn't profitable, but they'll still hook you up with a PC if you want. Or a Mac. Probably even a Solaris box.
I think there's a good chance Apple will license OS X to IBM for high end workstations, despite their policy and everything that's been said. PowerMacs don't have the features necessary to be used for high end workstation tasks in a lot of places (For example, lack of ECC memory. Yes, I know you've never had problems with Apple memory, but that doesn't make ECC optional for mission critical jobs.).
Apple's not going to make custom PowerMacs for niche markets, but IBM could and OS X would be a sweet edition for that.
A workstation edition of MacOS could easily command a much higher price, so Apple would keep their margin and expand into markets they can't touch right now.
"If OS X is the only real desktop alternative, nothing is stopping people from buying Macs you know."
For every "big mistake" that was going to kill Apple and didn't, there's a killer app that was going to take over the world and didn't. About the only world they've taken over is the portable music player market, and that's platform agnostic (even if some people switch as a result).
Hell that's what I thought the mini would be during that orgy of speculation in the week or two before the keynote. A small box that can leverage a large box for media purposes.
I also speculated that Apple might offer a downloadable movie service. Even if you have to start the download 1/2 an hour early to let it buffer, that's still better than going to Blockbuster.
The bandwidth costs would be hefty for Apple, and having some kind of p2p thing doesn't seem like their style even if they made it seamless and DRMed.
I dunno.
With Apple you can guess price points but you can't guess the crazy shit they will pull out of their asses, and you also can't guess the obvious Good Things that they will pass up.
Also, despite Oppenheimer's claims they won't be licensing MacOS X to OEMs I think there's a very good chance they'll license it to IBM for high end workstations that won't compete with PowerMacs.
I think it's a combination.
-The prices Dell gets for Intel processors and chipsets and other parts make AMD look expensive.
-Dealing with greater variety of processors and other parts increases overhead for manufacturing and support.
-Prejudice in the industry left over from earlier generations of AMD chips that weren't as solid.
-AMD chips are performing very well everywhere right now, but the difference is big enough for big customers to start switching mostly on big multi-processor systems, which is a tiny minority of all the computers out there (though the margins are nice).
Yeah, that was my point. I went to significant lengths to eliminate clumping and it still seems like it's there.
Using rand(3) would be part of the problem. It's known for having low randomness for the low order bits, at least older and less complicated implementations are. That is very likely the cause of the problem.
Shuffling is a lot easier when you have a big computer with lots of memory and all the libraries you could want.
It's not that it's impossible to do on an iPod, but it's quite conceivable that they never got around to it.
meh
/dev/urandom... the resulting playlists still seemed to have clumping. I dunno.
I wrote a little program to shuffle lines from stdin with drand48 seeded from
I really don't care enough to conduct anything approaching the kind of test it would take to verify that kind of thing.
Disclaimer: I have a shuffle
It is indeed treated as a standard USB mass storage device. There's a setting for how much space to reserve for user files. It's FAT32. I don't have any Windows machines but I can mount it on my Linux machine just fine. Can't seem to get it to charge on Linux though, I think it needs software on the host machine to charge. Hoepfully the gtkpod people will update soon.
Battery is rated at 12 hours but I seem to be able to get more, and apparently using AAC 128 kbps extends it. I have iTunes transcode my high bitrate MP3 files to AAC for this reason (also to increase the number of songs). There's generation loss, but the second generation files in the shuffle are never copied anywhere else so it's okay.
Well, it's really annoying when certain songs come up over and over while others go unplayed, especially if you like the ones that are missed.
I had a bad MP3 CD player (not an Apple product, as they've never made such a device) that would repeat some songs with a probability of maybe 1 in 20, and there were other songs that went unplayed unless I specifically picked them. With under a hundred songs and me going months between burning new disks, the probability of that is very low with a good pseudo-random function.
My iPod seems to favor some bands, but that could be just me. It's certainly not as bad as that piece of crap CP player.
The obvious thing to do is probably to gather a statistically significant number of trials and see what the distribution is then. Any bias would become apparent.
"Computers can't generate true random numbers (ok, at least I don't know of any current methods) but only pseudo random numbers. There's a precise mathematical description that gets you from one number to the next."
No shit.
But they can gather entropy from lots of sources. For example, the timing of user interactions.
It can continue to generate electricity at night because heat energy is stored at the bottom, and as the environment cools at night (as it does fairly quickly in cloudless environments like that) the stored heat can allow the cooler air to rise.