What it may mean is that SCO might charge Unix users who have moved to Linux from Unix but are still using a pair of old Unix ABI software libraries found in UnixWare and OpenServer to run Unix binaries on Linux.... On Linux, these libraries can be used to run some SCO Unix binaries.
.. When my old company was moving from SCO to linux, we experimented with the iBCS linux module to run some SCO OpenServer binaries. It basically worked for simple command-line progs but we could never get dynamically-linked X programs to work (because of the library dependencies). We ended up ditching all the SCO stuff.
I wonder if SCO wants to grab license fees only from people who copy SCO libraries onto their linux platforms, or even from people who run statically-linked SCO apps on Linux, which would be a different thing.
For one thing, Microsoft violated the law and was declared an illegal monopoly.
ahem. A legal monopoly, which then proceeded to use their monopolistic power illegally.
I think it would've been wrong for Judge Jackson to punish Microsoft harshly for "not knowing" it was a monopoly and their consequental conduct... had it not been so bloody stupid obvious that MS was indeed a monopoly.
You're right about Google not having been declared a monopoly. I don't think that by itself makes it "not" one. OTOH, monopolies are allowed to continue crushing competitors in the specific field they monopolize, so I don't see any (legal) problem with Google's behavior.
You will notice the difference between a game that runs at 30 stable FPS and one that runs at 60 stable FPS. The 24-30 FPS on TV and cinema is only perceived as fluid because of motion blur.
I *wondered* why I couldn't seem to focus on the background in The Two Towers' panoramic scenes until the camera stopped moving.
If everyone reused the standard APIs in an intelligent way instead of redesigning them incorrectly, they might actually get some code out the door.
The fun part is when the standard APIs really do suck, like select()/poll().
Then, real architecture questions need to be asked (do we use a bunch of threads instead? RT signals? epoll() where available? maybe a hybrid approach? How will it scale?). The answer varies depending on your platform and application, eg. can the various tasklets interfere w/each other. God help you if you're supposed to be multi-platform.
Be you theorist or code-cranker, you need to have some code architect skills and a deep understanding of the problem and platform(s) you're working on to do it right the first time. (As I'm painfully finding out.)
The arguments you mark as "economical" also apply to the Fibre Channel vs. SCSI question -- except that Fibre Channel solutions ended up being more expensive. I'm not sure if this is due to various development costs (FC really was intended to be a "revolution"), or simply because some manufacturers decided they could get away with it.
On x86, this works (at least for the calls I've tried). There are certain platforms (MIPS, and it's the only one I've dinked w/ besides x86) where no one ever bothered to get the do-interrupt-from-kernel functionality working:-(. Exporting the syscall table, exporting each asmlinkage function individually, or fixing that functionality are all possible solutions. The first one was the easiest (and should be faster than doing a software interrupt).
What would really be cool is if they had a kit that allowed one to right their own "code morphing" code on top of their vliw core
I thought TMTA's whole point was that they could completely redesign the core to be, I dunno, really-freakin'-large-IW and no apps would break -- because they all used the code-morphing interface. Giving direct access to the VLIW core would defeat that.
As I recall, NexGen (who was bought by AMD) did expose their pentium-class x86's RISC-like core. It had a register switch or something. Nobody I can recall made use of it, though.
But as long as we're dreaming, I'd think bigger. Like Bochs with multiple front-end instruction sets.
There are some notable kernel developers who would claim that changing an occurance of EXPORT_SYMBOL_GPL() to plain EXPORT_SYMBOL() (well, actually the alternative -- lying about MODULE_LICENSE) would be a violation of the DMCA. Not that these people are lawyers or even necessarily have a clue, but they are clearly hostile to the idea.
On a different tack, one technical objection to removing EXPORT_SYMBOL(sys_call_table) is that in addition to not being able to replace default system calls (which was always a bad idea), one can no longer call system calls from kernel-land that aren't specifically exported. That means there is kernel functionality available to userland that isn't available to modules. That is IMHO limiting and generally a Bad Idea.
I've tried the "ya know, this really needs changing, and here's a few reasons why..." approach. The response I've gotten was "No. You're an idiot. Your idea is stupid. We'll never do that. Go away."
Linus did this very thing to Andrea Arcangeli. Andrea's persistence eventually paid off.
I'm not saying Linus was right (and he's probably just as much of an asshole as he claims to be) but you can workaround the defects in the kernel developer foodchain if you keep plugging away and acquire better experience. The hard part of course is doing that instead of just saying "fuckit", which is what most sane people would do.
Re:Open Source is NOT the issue - it's the IMAGE
on
Largo Loving Linux
·
· Score: 1
I'm sorry, I don't have issues with most of your comment, but this suggestion:
The Penguin logo MUST go ASAP.
What?! That's your ronald mcdonald happy face right there!
I have started drinking, feeling that it is better to be passed out on the floor than gaming (it's cheaper, too.)
Is this a cry for help? It's not better. Don't do this. Simply put, every other problem of yours will get worse. Plus you'll have lots of new ones. Think about it.
I used to game a bit. I don't really anymore. Running linux full-time helps with that btw:) Find something healthy/constructive that you like to do, and get addicted to that instead. Some people call this a "hobby". You'll still be addicted, but at least your liver won't be fucked up.
I'd buy and old underperforming chip in 0.13u technology if it lets me get rid of the noisy fan.
I've seen cards with the three older chipsets mentioned; none of them, as I recall, had or needed a fan. Which isn't to say you could only buy fanless configurations.
If "doesn't need a fan" is your buying metric, a Matrox g450 or g550 would probably suit you (disclaimer: I own a g200, which is also fanless).
Now what's the best desktop (x86) *processor* I could get that doesn't need its own fan?
My last apartment (actually a co-op) had a local network that was constantly maxed out, and subsequently unusable, because somebody in disregard of the network AUP decided to run Napster, or download a DVD rip of The Matrix. Yeah, this was awhile back.
The same principle applies to those fellas uncapping their modems. It affects other people on the cable link. File trading is damn well "intrusive".
I don't get how everybody thinks cable uncappers should get off w/a slap on the wrist, while spammers should be shot. (Yes, it's because/.'ers are differently opinioned and unique. Just like everyone else.)
(*hurl*).. you threw that in on purpose right? That guy who played Evil Sting was kindof cool, the rest of the plot had approximately the depth of X-Men vs Street Fighter. My vote goes to Borg.
I'm more impressed by the true zero-copy send(2)s they claim to be able to do in -CURRENT, provided that the data meets certain alignment and size requirements, and that you manipulate the sndbuf properly. (I won't count zero-copy recv since it apparently only works on one card..)
zero-copy send() is more useful/generic than linux's sendfile() IMHO. To get sendfile(), you can just mmap() a file and send() it. It doesn't work so nicely the other way around.
Offtopic (was Re:CeBIT is better anyway)
on
The Last Comdex?
·
· Score: 0, Redundant
The problem with that sig is that it's infantile 'humor'... well, and that it's practically flamebait in itself.
Slashdot Mirror
Except for Linux (and the Hurd, which borrowed Linux's stack).
What it may mean is that SCO might charge Unix users who have moved to Linux from Unix but are still using a pair of old Unix ABI software libraries found in UnixWare and OpenServer to run Unix binaries on Linux. ... On Linux, these libraries can be used to run some SCO Unix binaries.
I wonder if SCO wants to grab license fees only from people who copy SCO libraries onto their linux platforms, or even from people who run statically-linked SCO apps on Linux, which would be a different thing.
ahem. A legal monopoly, which then proceeded to use their monopolistic power illegally.
I think it would've been wrong for Judge Jackson to punish Microsoft harshly for "not knowing" it was a monopoly and their consequental conduct... had it not been so bloody stupid obvious that MS was indeed a monopoly.
You're right about Google not having been declared a monopoly. I don't think that by itself makes it "not" one. OTOH, monopolies are allowed to continue crushing competitors in the specific field they monopolize, so I don't see any (legal) problem with Google's behavior.
I *wondered* why I couldn't seem to focus on the background in The Two Towers' panoramic scenes until the camera stopped moving.
gesundheit (bless you) :)
The fun part is when the standard APIs really do suck, like select()/poll().
Then, real architecture questions need to be asked (do we use a bunch of threads instead? RT signals? epoll() where available? maybe a hybrid approach? How will it scale?). The answer varies depending on your platform and application, eg. can the various tasklets interfere w/each other. God help you if you're supposed to be multi-platform.
Be you theorist or code-cranker, you need to have some code architect skills and a deep understanding of the problem and platform(s) you're working on to do it right the first time. (As I'm painfully finding out.)
I don't see what the problem would be with an earth-size moon orbiting a gas giant within the habitable zone ...
Ceti Alpha V
What the hell is wrong w/Mozilla? Other than that it did in fact take 4.something years to do .. that's a heck of a sacrifice.
That was Mark Twain, and it's timeless advice.
wtf? The article says the 'puter runs (some flavor of) AIX and Linux. Probably not sparc-based then. Are you talking about some other 'nodes'?
The arguments you mark as "economical" also apply to the Fibre Channel vs. SCSI question -- except that Fibre Channel solutions ended up being more expensive. I'm not sure if this is due to various development costs (FC really was intended to be a "revolution"), or simply because some manufacturers decided they could get away with it.
On x86, this works (at least for the calls I've tried). There are certain platforms (MIPS, and it's the only one I've dinked w/ besides x86) where no one ever bothered to get the do-interrupt-from-kernel functionality working :-(. Exporting the syscall table, exporting each asmlinkage function individually, or fixing that functionality are all possible solutions. The first one was the easiest (and should be faster than doing a software interrupt).
I thought TMTA's whole point was that they could completely redesign the core to be, I dunno, really-freakin'-large-IW and no apps would break -- because they all used the code-morphing interface. Giving direct access to the VLIW core would defeat that.
As I recall, NexGen (who was bought by AMD) did expose their pentium-class x86's RISC-like core. It had a register switch or something. Nobody I can recall made use of it, though.
But as long as we're dreaming, I'd think bigger. Like Bochs with multiple front-end instruction sets.
Also, I'd like a pony.
On a different tack, one technical objection to removing EXPORT_SYMBOL(sys_call_table) is that in addition to not being able to replace default system calls (which was always a bad idea), one can no longer call system calls from kernel-land that aren't specifically exported. That means there is kernel functionality available to userland that isn't available to modules. That is IMHO limiting and generally a Bad Idea.
Linus did this very thing to Andrea Arcangeli. Andrea's persistence eventually paid off.
I'm not saying Linus was right (and he's probably just as much of an asshole as he claims to be) but you can workaround the defects in the kernel developer foodchain if you keep plugging away and acquire better experience. The hard part of course is doing that instead of just saying "fuckit", which is what most sane people would do.
The Penguin logo MUST go ASAP.
What?! That's your ronald mcdonald happy face right there!
Is this a cry for help? It's not better. Don't do this. Simply put, every other problem of yours will get worse. Plus you'll have lots of new ones. Think about it.
I used to game a bit. I don't really anymore. Running linux full-time helps with that btw :) Find something healthy/constructive that you like to do, and get addicted to that instead. Some people call this a "hobby". You'll still be addicted, but at least your liver won't be fucked up.
I don't disagree with your claims. But talk is cheap, especially when it doesn't have a name behind it, AC.
I've seen cards with the three older chipsets mentioned; none of them, as I recall, had or needed a fan. Which isn't to say you could only buy fanless configurations.
If "doesn't need a fan" is your buying metric, a Matrox g450 or g550 would probably suit you (disclaimer: I own a g200, which is also fanless).
Now what's the best desktop (x86) *processor* I could get that doesn't need its own fan?
Edumacated readers will note that the parent post is a (well-done!) variation on the Tao of Programming. Nice job AC.
My last apartment (actually a co-op) had a local network that was constantly maxed out, and subsequently unusable, because somebody in disregard of the network AUP decided to run Napster, or download a DVD rip of The Matrix. Yeah, this was awhile back.
The same principle applies to those fellas uncapping their modems. It affects other people on the cable link. File trading is damn well "intrusive".
I don't get how everybody thinks cable uncappers should get off w/a slap on the wrist, while spammers should be shot. (Yes, it's because /.'ers are differently opinioned and unique. Just like everyone else.)
My props to Mr. McDowell. It's too bad that he and Mr. Stewart could not save Generations.
(*hurl*) .. you threw that in on purpose right? That guy who played Evil Sting was kindof cool, the rest of the plot had approximately the depth of X-Men vs Street Fighter. My vote goes to Borg.
zero-copy send() is more useful/generic than linux's sendfile() IMHO. To get sendfile(), you can just mmap() a file and send() it. It doesn't work so nicely the other way around.
The problem with that sig is that it's infantile 'humor' ... well, and that it's practically flamebait in itself.