Well, see, here's how the Open Source model works. You let the whole world see your code, and then anybody who cares can make changes, and then anybody who cares about the changes can use them (including you, which is why you opened up the code in the first place - so you could get other people's changes).
Yeah, I know, that's a very high-level view, and it ignores all the license issues and whatnot, but that's how the social dynamics of it work.
So if Microsoft wants to do that (with any project), that's fine, and if they pick something people care about, they'll get a lot of good code out. But...
You have to open up the whole project, or else you get into motivational problems (the parent poster's "what do I get out of it?"), scope problems ("I really want to improve this other area, but they didn't give me the source to that"), and procedural problems ("how do I test my changes when I can't compile the whole project because I don't have the source to it all?").
You have to make people's changes easily available to other people, so that somebody can improve somebody else's changes.
Related to the previous point, you need somebody in charge of accepting/refusing changes, so that there is a single "current" or "official" state of the project, so that it doesn't die the death of a thousand forks. And that person in charge has to be trusted to Do The Right Thing, or people won't be willing to work on the project. ("The right thing" almost always means "make decisions on ruthlessly technical grounds" - no favorites, not even Microsoft.) Some humility and a sense of humor in this person will also help. See Linus for a great example.
If Microsoft is willing to do that, great. But half measures will earn a less-than-half response from the community, because the social dynamics are all wrong.
No, SCO is not "ceding its ownership rights of the code by distributing it". SCO is giving permission for others to use it under the GPL. This means SCO gives up some control that it had, but that's not the same as giving up ownership.
Yes, it is - as long as you stick to the POSIX specification. As the article points out, as soon as you go past that, Solaris isn't open any more, and neither are any of the other UNIXes. It's not open by the parent's definition (which I like, BTW) because there isn't any open standard for the non-POSIX parts of Solaris.
Part of the point of the article is that there is a lot of stuff in the "non-POSIX" part of Solaris. And if you use it, you're stuck with Solaris, and so Solaris isn't open to you.
I agree with the parent's definition. But Solaris isn't as good a fit as he thinks.
BTW, it's possible for open source to be closed in terms of open standards. (Of course, you could always re-engineer a new standard from the source, but then you run into one of the usual problems with standards - there are so many of them to choose from.)
Um, yeah, actually we are doing it because we can. See, we're tired of Microsoft "leveraging the installed base." That means Microsoft trying to force us to do something we don't want to do. We're tired of Microsoft trying to trap us in a situation where we have to keep sending them money in order to keep our computers working. We're leaving this mess because we can.
And yeah, I personally am still on Windows. But I believe I understand what's driving people to switch. (Why haven't I switched yet? Educational software for the kids.)
It isn't "security through obscurity". It's "guaranteed worst-case response time through using a real real-time (not just multi-tasking) OS". Windows is multi-tasking, but it isn't a hard real-time kernel. AFAIK, Linux isn't either.
Sure, you can modify the Linux kernel. But if you do, you don't have a million man-hours on your modifications.
The distinction about "off the shelf" is between that and "roll your own". Off the shelf would include vxWorks, Green Hills, and pDos and OS-9 (if they are still around), and probably a few others.
I'm most familiar with vxWorks, so I'll talk about that one. If you don't need, say, TCP/IP, you can simply take it out. Your memory footprint just went down. Don't need memory management? Don't put it in. Don't need disk support? Remove it. Need to initialize something before the kernel starts time-slicing? They've got a standard hook for that - no hacks needed. Want to run on a PowerPC chip? Supported. Motorola ColdFire? Ditto. MIPS? Ditto.
Back to quality: The core code of an embedded OS has been beat to death in that environment, and proven rock-solid. No "if the wrong interrupt comes at just the wrong time, it goes off into an extended thrashing session for several seconds". Their customers simply won't put up with the kind of semi-broken behavior that Windows exhibits all too frequently.
That would actually be good. Hack the vote, not to throw the election one way or the other, but to clearly show the public what the problem is. If Mickey Mouse is elected president, that would illustrate the issue nicely, in a way that the public can grasp.
No, it's not just SCO's word that such an agreement exists between SCO and IBM. Groklaw has a copy of the agreement posted for all to see.
People on Groklaw have also noted that the agreement says that SCO can't take the Project Monterrey stuff and put it on a non-Intel processor. The agreement does not say that IBM can't put the code on another processor. (Note that I did not verify this wording in the agreement myself, so if you don't want third-hand info, head to Groklaw, read the agreement, and make up your own mind.)
Wake me up when SCO actually says it in court. What they say to the press is so disconnected from reality that I refuse to bother worrying about "what if" this one happens to actually be true.
I can type faster than I can write longhand. At least, I think I can - I haven't actually tried racing one method against the other. I'm pretty sure, though, that I can type faster than I can write plus correct all the errors in the handwriting recognition software.
On the other hand, I know that I can talk faster than I can type. But if I used speach recognition, that sentence would have come out "On the other hand, uh, I'm pretty sure, no, I know that I can talk faster than I can type." That's without the speech software having any problems doing the translation.
Bottom line: Typing is higher bandwidth than any other currently available input method.
About the Democratic convention
on
We the Media
·
· Score: 2, Insightful
"The Charge Of The Blog Brigade" (with apologies to Alfred, Lord Tennyson)
Theirs not to wonder why Theirs but to blog and die Into the valley of hype rode the six hundred.
Boredom to the right of them, Boredom to the left of them, Boredom in the front of them, Into the valley of hype rode the six hundred (bloggers).
Back to actual commentary: Of course the bloggers at the convention said nothing of consequence. Nothing of consequence happened at the convention. It never does. Nothing will happen at the Republican Convention either.
Conventions used to be about deciding who your candidate was going to be, and what your platform was. But these days, we know who the candidate will be before the convention starts, and the "platforms" are designed to sound as good as possible without actually saying much that is concrete.
The result is that conventions generate no actual news. What, Kerry was nominated? Really??? Wow, that's really news!
So I'm not sure that the Democratic Convention is a good proof that blogs don't really cut it as the new news media. If there's no news, the professional evening news doesn't say much either.
I think you misunderstand. Daxx's point is that people won't care as long as the issue is abstract. When it starts to hit them personally, there will be action.
An idealist would say that people should care anyway, even if they aren't personally affected. A pessimist (the parent) would say that people will never care, no matter what. A realist says people are motivated most strongly by self-interest. Me, I agree with the idealist about what should be, and with the realist about what is;-)
A cynical realist says that the masses care only when it limits their bread and circuses. But inability to copy DVDs actually does cut into that...
No, it wasn't just about a blow job. It was about selling the country to the Chinese (remember Lippo), about a guy whou could argue about what the definition of "is" is rather than just tell the truth (to a Congressional inquiry, no less), about renting out the Lincoln bedroom for campaign contributions, and I'm sure I'm forgetting a number of other gems.
Free clue: Military intelligence is always really, really hard. See, Saddam wasn't sitting there trying to make it easy for the CIA to find out exactly what Iraq did or did not have. It appears that Saddam was trying to look like he did in fact have WMD, perhaps to tweak Bush, perhaps to keep Iran from getting any ideas, perhaps just because Saddam was an idiot.
The point is, military intelligence is always incomplete. Worse, it's often overwhelmingly incomplete - it's a huge amount of information that still doesn't give you the big picture correctly. There's almost always the pieces to put together the big picture, but they are buried in so much other stuff that figuring out which are the right pieces is almost impossible.
So you make your best guess about what's going on (they call them intelligence "estimates" for a reason), and you go with it. You have to.
The open source community is exactly that, a community. It's not a company, it's not just an idealogy, it's not just a methodology, it's a community.
The community itself is the prior art database. We've got all these people who have worked all kinds of places and seen all kinds of software and coding techniques. We don't want to write it all down, but if any particular software patent ever starts biting you, just put it out for comment on the net (say, Slashdot, Groklaw, probably a few other places as well). You'll get lots of responses. Most of them won't help, but one or two may be close enough to count as prior art. If the prior art is out there, this seems to me to be the most likely way to find it.
Creating a database? I've been programming for twenty years (all right, nineteen and a half). If I were to try to write down every technique I've seen so far in my career, I'd forget half of them, and I'd quit from boredom long before I even wrote down the half I remembered. This doesn't mean that I've really forgotten the other half, though. I still remember them if someone asks...
Yeah, I kind of see open source as a bunch of people that you do not want to get into a patent fight with. They care passionately, they have way too many people that remember lots of things that could be prior art, and they know how to network with each other. (See Groklaw for an example.) In a serious patent fight against open source, the patent holder would stand to see patents (or individual claims of patents) ruled invalid due to prior art, and the software swiftly re-written to not infringe on the rest. Best to attack open source only with patents that you don't care about losing.
Not to say that it wouldn't be a pain in the Tux...
Oh, moving some stuff to user mode? Well, um, better late and half-baked than never and not at all?
Seriously, putting stuff in the kernel that should have been in user space is one of the more serious architectural botches in Windows. It has caused massive stability problems. Now it seems that Microsoft is recognizing this, and is starting to undo it. (What they need is to completely undo it, but they have to start somewhere. What they don't get to will continue to bite them until they do.)
Well, Linux is fine for embedded systems. There's two fronts where real "embedded" OSes may be ahead, though: Hard real-time and memory footprint. A real-time OS like vxWorks probably beats Linux hollow on things like worst-case latency, and possibly on context-switch time as well. Also, vxWorks lets you only pull in the parts you need. Linux does also, but I suspect that vxWorks can take much less memory than Linux. As time goes on, of course, this becomes less relevant, as memory gets cheaper.
As to whether Windows can run without an MMU: It originally ran (badly) on a 286, which I believe did not have an MMU. Whether modern Windows can run without an MMU is a different question;-)
After reading several of your posts on this topic, I have to say that you just don't get it.
I have something that I want a computer to do. Maybe it's to render frames for an animated movie. Maybe it's to run a database. But it's not to run a particular OS. Nobody (well, almost nobody) in the business world is religious about the OS, they just want to get their work done.
And very often, what I want the computer to do is to run some custom application. In other words, regaredless of what the OS is, the business still has to hire people to support the custom app. (Note, however, that very few of these apps require custom kernels in Linux. Also note that those that do, require things that would require custom kernels in Windows, too, except that the Windows license doesn't let you do that. In other words, if you have to hire additional devs to maintain your kernel, it probably would not even be possible to do the same thing on Windows. But I digress.)
Here's the macroeconomic point: This is not a zero-sum game. Switching to an OS with lower total cost of ownership makes things economically possible that were not possible before. In other words, the whole economy grows as people stop wasting so much money on OSes. That means a net increase in jobs.
And yes, this does apply to Linux vs. Windows. You need to replace your Windows admins with Linux admins, but one Linux admin can take care of more boxes, so you don't need as many admins. You can spend less on admins and more on what your business is really trying to do. This can help your business grow (meaning you can employ more people).
Slashdot Mirror
Yeah, I know, that's a very high-level view, and it ignores all the license issues and whatnot, but that's how the social dynamics of it work.
So if Microsoft wants to do that (with any project), that's fine, and if they pick something people care about, they'll get a lot of good code out. But...
You have to open up the whole project, or else you get into motivational problems (the parent poster's "what do I get out of it?"), scope problems ("I really want to improve this other area, but they didn't give me the source to that"), and procedural problems ("how do I test my changes when I can't compile the whole project because I don't have the source to it all?").
You have to make people's changes easily available to other people, so that somebody can improve somebody else's changes.
Related to the previous point, you need somebody in charge of accepting/refusing changes, so that there is a single "current" or "official" state of the project, so that it doesn't die the death of a thousand forks. And that person in charge has to be trusted to Do The Right Thing, or people won't be willing to work on the project. ("The right thing" almost always means "make decisions on ruthlessly technical grounds" - no favorites, not even Microsoft.) Some humility and a sense of humor in this person will also help. See Linus for a great example.
If Microsoft is willing to do that, great. But half measures will earn a less-than-half response from the community, because the social dynamics are all wrong.
IANAL, but I read both Slashdot and Groklaw...
Well, not really. But if the site goes down, or your net connection does, your apps might as well be blue-screened...
Yes, it is - as long as you stick to the POSIX specification. As the article points out, as soon as you go past that, Solaris isn't open any more, and neither are any of the other UNIXes. It's not open by the parent's definition (which I like, BTW) because there isn't any open standard for the non-POSIX parts of Solaris.
Part of the point of the article is that there is a lot of stuff in the "non-POSIX" part of Solaris. And if you use it, you're stuck with Solaris, and so Solaris isn't open to you.
I agree with the parent's definition. But Solaris isn't as good a fit as he thinks.
BTW, it's possible for open source to be closed in terms of open standards. (Of course, you could always re-engineer a new standard from the source, but then you run into one of the usual problems with standards - there are so many of them to choose from.)
Only if you don't have to take any apps with you...
There's only one?
And yeah, I personally am still on Windows. But I believe I understand what's driving people to switch. (Why haven't I switched yet? Educational software for the kids.)
Sure, you can modify the Linux kernel. But if you do, you don't have a million man-hours on your modifications.
The distinction about "off the shelf" is between that and "roll your own". Off the shelf would include vxWorks, Green Hills, and pDos and OS-9 (if they are still around), and probably a few others.
I'm most familiar with vxWorks, so I'll talk about that one. If you don't need, say, TCP/IP, you can simply take it out. Your memory footprint just went down. Don't need memory management? Don't put it in. Don't need disk support? Remove it. Need to initialize something before the kernel starts time-slicing? They've got a standard hook for that - no hacks needed. Want to run on a PowerPC chip? Supported. Motorola ColdFire? Ditto. MIPS? Ditto.
Back to quality: The core code of an embedded OS has been beat to death in that environment, and proven rock-solid. No "if the wrong interrupt comes at just the wrong time, it goes off into an extended thrashing session for several seconds". Their customers simply won't put up with the kind of semi-broken behavior that Windows exhibits all too frequently.
That would actually be good. Hack the vote, not to throw the election one way or the other, but to clearly show the public what the problem is. If Mickey Mouse is elected president, that would illustrate the issue nicely, in a way that the public can grasp.
But you'd better not get caught...
For real justice, try writing penguin.com (as was mentioned in the interview with Katie). Or, for bonus points, try microsoft.com
Problem is, paying $75 to call yourself a pilot may let you fly around, but it doesn't mean you can land...
Yes. Now go actually read that section. It does not limit IBM to any specific architecture - unlike the restriction placed on SCO.
People on Groklaw have also noted that the agreement says that SCO can't take the Project Monterrey stuff and put it on a non-Intel processor. The agreement does not say that IBM can't put the code on another processor. (Note that I did not verify this wording in the agreement myself, so if you don't want third-hand info, head to Groklaw, read the agreement, and make up your own mind.)
Wake me up when SCO actually says it in court. What they say to the press is so disconnected from reality that I refuse to bother worrying about "what if" this one happens to actually be true.
Their Lanham Act counterclaims in the lawsuit that SCO filed are directly referring to this type of behavior.
I can type faster than I can write longhand. At least, I think I can - I haven't actually tried racing one method against the other. I'm pretty sure, though, that I can type faster than I can write plus correct all the errors in the handwriting recognition software.
On the other hand, I know that I can talk faster than I can type. But if I used speach recognition, that sentence would have come out "On the other hand, uh, I'm pretty sure, no, I know that I can talk faster than I can type." That's without the speech software having any problems doing the translation.
Bottom line: Typing is higher bandwidth than any other currently available input method.
"The Charge Of The Blog Brigade"
(with apologies to Alfred, Lord Tennyson)
Theirs not to wonder why
Theirs but to blog and die
Into the valley of hype rode the six hundred.
Boredom to the right of them,
Boredom to the left of them,
Boredom in the front of them,
Into the valley of hype rode the six hundred (bloggers).
Back to actual commentary: Of course the bloggers at the convention said nothing of consequence. Nothing of consequence happened at the convention. It never does. Nothing will happen at the Republican Convention either.
Conventions used to be about deciding who your candidate was going to be, and what your platform was. But these days, we know who the candidate will be before the convention starts, and the "platforms" are designed to sound as good as possible without actually saying much that is concrete.
The result is that conventions generate no actual news. What, Kerry was nominated? Really??? Wow, that's really news!
So I'm not sure that the Democratic Convention is a good proof that blogs don't really cut it as the new news media. If there's no news, the professional evening news doesn't say much either.
An idealist would say that people should care anyway, even if they aren't personally affected. A pessimist (the parent) would say that people will never care, no matter what. A realist says people are motivated most strongly by self-interest. Me, I agree with the idealist about what should be, and with the realist about what is ;-)
A cynical realist says that the masses care only when it limits their bread and circuses. But inability to copy DVDs actually does cut into that...
No, it wasn't just about a blow job. It was about selling the country to the Chinese (remember Lippo), about a guy whou could argue about what the definition of "is" is rather than just tell the truth (to a Congressional inquiry, no less), about renting out the Lincoln bedroom for campaign contributions, and I'm sure I'm forgetting a number of other gems.
The point is, military intelligence is always incomplete. Worse, it's often overwhelmingly incomplete - it's a huge amount of information that still doesn't give you the big picture correctly. There's almost always the pieces to put together the big picture, but they are buried in so much other stuff that figuring out which are the right pieces is almost impossible.
So you make your best guess about what's going on (they call them intelligence "estimates" for a reason), and you go with it. You have to.
The open source community is exactly that, a community. It's not a company, it's not just an idealogy, it's not just a methodology, it's a community.
The community itself is the prior art database. We've got all these people who have worked all kinds of places and seen all kinds of software and coding techniques. We don't want to write it all down, but if any particular software patent ever starts biting you, just put it out for comment on the net (say, Slashdot, Groklaw, probably a few other places as well). You'll get lots of responses. Most of them won't help, but one or two may be close enough to count as prior art. If the prior art is out there, this seems to me to be the most likely way to find it.
Creating a database? I've been programming for twenty years (all right, nineteen and a half). If I were to try to write down every technique I've seen so far in my career, I'd forget half of them, and I'd quit from boredom long before I even wrote down the half I remembered. This doesn't mean that I've really forgotten the other half, though. I still remember them if someone asks...
Not to say that it wouldn't be a pain in the Tux...
Oh, moving some stuff to user mode? Well, um, better late and half-baked than never and not at all?
Seriously, putting stuff in the kernel that should have been in user space is one of the more serious architectural botches in Windows. It has caused massive stability problems. Now it seems that Microsoft is recognizing this, and is starting to undo it. (What they need is to completely undo it, but they have to start somewhere. What they don't get to will continue to bite them until they do.)
As to whether Windows can run without an MMU: It originally ran (badly) on a 286, which I believe did not have an MMU. Whether modern Windows can run without an MMU is a different question ;-)
I have something that I want a computer to do. Maybe it's to render frames for an animated movie. Maybe it's to run a database. But it's not to run a particular OS. Nobody (well, almost nobody) in the business world is religious about the OS, they just want to get their work done.
And very often, what I want the computer to do is to run some custom application. In other words, regaredless of what the OS is, the business still has to hire people to support the custom app. (Note, however, that very few of these apps require custom kernels in Linux. Also note that those that do, require things that would require custom kernels in Windows, too, except that the Windows license doesn't let you do that. In other words, if you have to hire additional devs to maintain your kernel, it probably would not even be possible to do the same thing on Windows. But I digress.)
Here's the macroeconomic point: This is not a zero-sum game. Switching to an OS with lower total cost of ownership makes things economically possible that were not possible before. In other words, the whole economy grows as people stop wasting so much money on OSes. That means a net increase in jobs.
And yes, this does apply to Linux vs. Windows. You need to replace your Windows admins with Linux admins, but one Linux admin can take care of more boxes, so you don't need as many admins. You can spend less on admins and more on what your business is really trying to do. This can help your business grow (meaning you can employ more people).