The problem with mp3.com was the artists had to sign over the rights to any songs mp3.com sold under the pay per play program. The same ass fucking took place just with a different entity behind the intense reaming. Maybe mp3.com acted better and didn't dick the artist around as much as one of the big four but the artist still had to fork over his or her rights to the work. For some I bet it was awesome to sell at least something without some uber-tough record contract squeezing out your creative spirit.
Re:Ok, so HURD is a microkernal os...
on
Hurd: H2 CD Images
·
· Score: 2
Performance of microkernels is definitely debatable. Microkernels on systems with slow context switching or a low number of active processes (like the PPC 603) perform rather badly because of a generally slow system and the required overhead of their design. However once you stick them on faster and more capable hardware you start to definitely see improcements. Monolithic kernels though run well in the environment where you need a little slimmer code because you're lacking speed/ability. Windows NT is still a microkernel design though just about everything runs in serverland as Win32 or userland as extensions of Win32. OSX is another good example of a microkernel design, one people seem to get confused. It is primarily a Mach kernel that runs a BSD 4.4 Lite based server and thus is compatible with BSD 4.4 Lite compatible software and functions. You can drop the BSD subsystem out of OSX and it will still work fine as the MacOS components based on Carbon and Cocoa don't require the BSD system at all.
Without the GNU project Linux as an OS would simply not exist. There MIGHT be a 1.x kernel floating around right now had there need no GNU project to actually do something with the kernel. Where do you think most of the shit redhat packages as their "base" system comes from? They sure as hell didn't write it themselves. It's also pretty retarded to say "why use HURD when there's Linux" because by the same logic you can say "why use Linux when there is [insert OS here]". How come Linux users are so quick to forget all of the stuff that had to happen for Linux to even exist. Shit man one wayward sperm and ther'd be no Linus to write Linux.
Re:Ok, so HURD is a microkernal os...
on
Hurd: H2 CD Images
·
· Score: 3, Informative
IANAKEOAS ( I am not a kernel expert of any sort) but microkernels differ from monolithic kernels is many ways other than hardware drivers running as services. They only pass messages between components meaning EVERYTHING is a server. Want to run Linux programs on a Mach kernel? Just run a Linux server and the software will run just fine. A good example of this is MkLinux and Windows NT. Windows NT runs a Win32 as a server but by the same token can run a POSIX server so POSIX compliant software also runs on it. MkLinux is a Mach kernel running a Linux server to run Linux software. Microkernels in theory are platform agnostic as they only pass messages between various servers. Most people except Linus Tovalds really dig microkernels and have put alot of work into them. Another advantage is you can remain platform agnostic in design yet run servers specific to hardware you're running on. You can go from running on a StrongARM system with 16 megs of RAM to a 32 processor x86 system using a majority of the same code. Notable microkernel based OSes include Windows NT, MacOS X, and of course HURD.
You're ignoring just about everything I said, that being big shared memory systems are better than clustered systems for particular jobs. If you want to serve half a gillion web pages or run a bunch of copies of FLUENT buy the Athlon cluster. Same with a big renderfarm doing batch rendering. If you want a realtime system you need a real computer. You said it yourself, you can malloc just about the entire system's memory for a single process. If say you need to run FLUENT with really big datasets clustering may or may not give you a price performance advantage. Instead of FLUENT I was thinking more along the lines of realtime video where an Onyx is feeding several channels of video each with specific content.
It seems to piss you off that somebody might suggest clustered systems are not the end all be all of computing. You seem to assume I know shit about Origin/Onyx systems when in fact I've read several essays written by SGI's engineers about them. I've also used older Origin systems (pre 3k systems) as well as lots of O2s doing realtime video work. I know the history of the XIO switching and the advent of the ccNUMA stuff. This doesn't make me an engineer but I'm also not talking out of my ass. I've seen plenty of specs for big and fancy Ethernet based Athlon clusters but they are dumb renderfarms. Its fine if you can break down some big problem into a bunch of easily digested packets for a cluster to munch on, in fact that's cool in my book if you can.
The P4 and Athlon beat out most other processors if compared CPU to CPU but after that they fail to scale. Adding N processors to a Ethernet based cluster does not scale well in terms of bandwidth and eventually you end up with diminishing returns and lots of wasted cycles because your cluster just doesn't feed its processors with enough data. It certainly won't process anything near realtime like a big iron system will. You should know this from experience and not ignore the failing of clusters in several cases.
Irix works well on the hardware it was written for. The original post claimed Irix was a POS and Linux could somehow magically be made for fill its role. I contend that Linux cannot and will not be made to fit this role because Irix has become too well adapted to its task. Trying to do the same with Linux would be damn near impossible if your goal was to make a Linux system that looks like a normal Linux system yet is running a 512 processor behemoth. I also didn't say anyone bought 32 processor machines to write code, they buy them to run code. Specifically code they've written. Shit man how many of your customers came to you saying "we've got this Fortran program we need to run in X amount of time, what do you suggest"? If the answer pisses YOU off then tough. Don't get your panties in a bunch because somebody didn't pee their pants when somebody said the word cluster.
I was just at the local mall earlier this afternoon. Whilst looking in Software Ect. I asked if I could play the XBox demo they had in the front window which had bee nturned off. The reason it was turned off is it had overheated a couple days before Christmas and they've yet to get a box to replace it with. Unless Rob lives in a universe parallel to my own with slightly different thermodynamics I don't think his comment was out of touch with reality.
Uh..with the ccNUMA structure of Origins a bad DIMM doesn't do much more than you're minus that many megs of memory overall. It is a switched architecture, the failure of one part of the system doesn't mean the whole thing is dead. If a processor module dies you just need to replace it. Have you seen the insides of C-Bricks? Unplug, unhook, slide out, slide in, hook, plug, done. You also make it seem like Irix is somehow a shitty OS on SGI hardware. Wow I want to spend a cool million of a huge Onyx and try to port Linux to it. Right. As for code being written for distributed machines that's just bullshit. Most if not everybody who's buying a system with more than 32 processors is writing their own software to run on it. Saying they are going to buy a MPP machine and partition it to run as a cluster is ridiculous. That is another point you are missing, Onyx and origin systems are NOT SMP systems, they are MPP (massively parallel processor) systems. SMP is where a couple processors populate a bus or have crossbar connects to one another. An individual C-Bick in an Origin might be considered an SMP system but the Origin itself is not. I'd really love to see you build a cluster than can do even half the things an Onyx system can. Renderfarms are just a bunch of retarded nodes that take individual frames and render them which is easily clustered. You don't buy an Onyx just to batch render a bunch of video frames. Your final point just doesn't make sense. When isn't an application based on the amount of memory and IO of a system? Who cares if an Athlon has a higher SPECfp score than a R12k. Clusters can't manipulate enormous datasets in realtime like MPP systems can.
Re:Eh? How can they get away with selling that?
on
HP-LX 1.0 Secure Linux
·
· Score: 4, Funny
Charging 3000$ for the CD set means that 99% of the jackasses who would use the GPL in order to buy something and then turn around and release it for free can't afford it while the 1% that can have to pay a pretty penny to be jackasses. I can pretty much assure you some jackass Linux zealot with no understanding on the GPL is sitting in his bedroom right now trying to figure out how he can raise 3k so he can be a folk hero by releasing the code an evil company is keeping secret. At the very least HP is giving some idiot something to do.
What the fuck are you talking about? Do you know anything about SGI hardware or software?
1. Saying SGI boxes do not scale completely invalidates any thought you've ever had in your life. Have you ever seen an Onyx/Origin system before? You add a brick and have twice the processing horse power or add 10 PCI (or any periphrial connection you ca nthink of) slots to do anything you want. You don't even have to turn the system off to add anything to it. Origins scale from 2 processor to 128 processor systems at the drop of a fucking hat. Using the Indy and O2 for your examples is just jackassery. Neither of which were designed for easy upgrades. Later revisions of the O2 when the Octane was released were easy as could be to upgrade. You turn the system off and slide out a processor modules and replace it with a new one. Voila.
2. Clusters fucking suck. Jesus get it through your Beowulf worshiping head that clusters are not the answer to all computation problems. LANL might build a bix cluster of PC systems to solve some embarrassingly redundant set of equations big fucking whoop. The big SGI systems are cache coherent and have messaging latencies measured in nanoseconds. A processor on one end of a 512 processor machine can talk to another processor with similar latency that a PC based system can talk to its own RAM. It is trivial to write software that spawns processes onto multiple nodes in a cluster, ccNUMA systems run as if their far flunt components were a single machine.
3. Linux cannot do that. If it COULD do that it would no longer be Linux and certainly would not be portable outside of SGI's hardware. Do you seriously think you can get good performance out of off the shelf software? Hell no. Optimizing software and good drivers doesn't mean anything. You've bitten far too deeply into Linux hype if you think it is even a contender to the grand daddy Unix systems.
Bringing up NewTek and ReelMagic cards is just fucking ridiculous. Wow you can have a hardware video processor that uses a PCI bus. That's rad...until you've seen an Origin in action running more than a dozen pipelines feeding 8 display channels. Try that with a PCI bus and see what happens. Nobody buys SGI for the logo.
A useful link would be here over at LinuxDevices.com. The board I might suggest to you in the Bitsy with its PCMCIA slot on board. With 16MB of RAm and 32MB of Flash ROM you'd have plenty of space to stick some software to run your car. Grab an off the shelf 802.11 PCMCIA card you can find drivers for and you've got it networked. The board also has USB which works for a camera and total of 21 digital IOs that you can use to control the components of the RC car. A bonus is the card also has audio in and out so you can turn your little car into a little talking robotic badass.
Strangely people talk about passive cooling in tower cases. Passive cooling would work much much better in a desktop chassis than in a tower. Reason being you could take a nice sized heat sink and stick it on your processor with plenty of headroom over it. With enough intake slits near the processor convetion would work pretty well for the most part. Heat sinks are perpendicular to the plane of the processor but making the plane of the processor perpendicular to the direction air is going to naturally flow ends up fucking up your cooling. More than a dozen models of Macs were designed this way and coupled with the fact the 603 disappated about 5 watts at the most made for a pretty quiet design overall, my PowerBook uses the same concept but instead of a heat sink it conducts heat into my lap. If you took a decent desktop chassis with a Tualatin P3, Celeron, or Via C3 you could probably get away with entirely passive cooling on the processor itself. As for the other noisy parts thats somebody else's problem.
Linux will never be like OSX because all of its zealot users are too anal. They clamour about choice like it's going out of style. With MacOS you write a program using one of a small number of APIs (Cocoa and Carbon specifically). If a program is written for OSX it just works. Linux folks want 20 different graphics toolkits. It is a natural tenent of the whole Bazaar concept that a good portion of all work overlaps in function. You're never going to see something like OSX come out of any completely open source project. The skill might be there but then to are the egos which make everybody go off and do their own thing.
What the fuck is this guy smoking? Like so many others have said this ought to be 10 pipe dreams for 2002. Not one thing he said made much sense at all and in general sounds like a half baked article he came up with at 6am to meet a 7am deadline.
1.) Purple twinky induced fantasy.
2.) Yet more fantasy, this time assuming users of Linux GUIs will be able to paste text between different applications written by different people. Back when there was growing dissatisfaction about IBM's licensing yet they are still the biggest computer company ever.
3.) This has been said for the past three years and has yet to happen. Why? Apps developed to scratch an itch are often not too broad in scope and have little intention of starting a paradigm. Apps intended to replace closed source counterparts rarely if ever achieve said couterpart's functionality. You end up with a system that doesn't talk to itself with a quarter of the features you could get for paying for something.
4.) People woken up from being cryogenically frozen for the past year could fucking tell you this.
5.) Right. Do you know what sort of systems REALLY classified data is stored on? Probably not. Three letter agencies don't exactly order their super secret computer systems from Dell with Windows 98 installed on them.
6.) No. That is just retarded. Unless several dozens apps somehow get ported to Linux magically this will not happen. No one is going to prepackage and OS they can't sell software for. Case in point BeOS.
7.) Maybe. Leo Laporte likes Linux but I've never heard him actually say anything he has really used it for. Wow you can replace a small handful of Windows programs with it that is sure to impress alot of people. A Linux TV show would be like a live action freshmeat.
8.) What the fuck? That is the most ridiculous thing in the entire article which in itself is notable. AOL on Linux would be like putting a vinyl interior in an Abrams tanks. It's asthetic vinyl...fitting to a fucking tank.
9.) *bong sound*
10.) see 9
It'd be damn near impossible to sell Linux to 99% of the people buying computers from retail sources. Even if you pushed it with millions of dollars in marketing you couldn't sell it to most people. If they couldn't call up a single number to get support if they have problems they won't buy the computer. It is that simple. No one wants to be refered to newsgroups and some crusty HOW-TOs when they run into a snag trying to get on the internet or when a program won't open because of some obscure library being in the wrong place. People are definitely not going to buy a system when they ask if they can edit some home movies or play video games and get laughed at by a sales person. Where's Linux's capability at? It doesn't matter if it can run for a year without rebooting. If there's nothing to use for a year for the retail computer buying public nobody's going to buy it.
Uh...the genetic difference between man and chimpanzee is about 1.6% which means we share 98.4% of our genetic material with chimpanzees. That 1.6% differences code all the phentotypes that makes us different from chimpanzees as well as our similarities.
When they were building the first PC they already had manufacturing rights to the 8088/8086 because they used the 8086 in the Displaywriter Intelligent Typewriter. Since they could already produce the chip it fit in well with their "off the shelf" component structure of the PC.
256 levels of contrast ratio is not nearly enough to match what the human eye can see. Colour negatives come far closer to what the eye can distinguish. Also there are only a very very small handful of cameras that offer more than 8 bits of quantization and these all cost more than 1000$ and then a small portion of those have interchangable lenses and the ability to do long term exposures. If you'd be so kind as to be specific as to why my price/performance assertion is bogus I'd really enjoy that. I'd love to see how a 300$ digital camera that takes pictures equivilent to a disposable Kodak camera at best beats even a crappy SLR camera. Maybe as you move up the rungs of camera quality high priced digital cameras might come out close to cost effective when compared to SLR camera but I really don't see it. I think I'd take my SLRs and my film scanned by professional scanning labs over a 5MP Canon until I scan upwards of 70 rolls a year which I don't happen to.
What about anything I said about bit depth is wrong? Besides I answered the question the guy asked and that was whether digital camers right now really compare to film which they don't. Price/performance wise they are completely uneconomical unless you have a point and shoot 35mm and take more than 40 rolls of film a year. People with those sorts of camers are lucky to develop 10 rolls a year. Here you you will be better informed read up. He says the exact same thing as me and is an admittedly better photographer. Fucking smug people.
Statistically all statistics lie in some form. Journalists statistically don't have any degrees besides journalism. Few journalists statistically know more about statistics than statisticians.
My point? You can infer anything from statistics, thats like the first chapter of a statistics textbook. So web eyeballs have narrowed their focus to a smaller number of websites in a given amount of time, big deal. If you go a little farther back in time you'll see the exact same thing as today, a majority of web users visited a handful of websites. Why? Lots of reasons. The biggest is only a smaller number of websites offered content the majority of web users even wanted to look at. Then there was a boom of websites that all offered the same thing packaged a little differently, some put blue bows on their piles of shit whilst others wrapped their shit in red bows. People liked the red bows more and thus now most of the blue bow sites are gone. Before the boom there were a handful of sites because no one thought much of the internet, now there are a handful of sites because people overvalued the internet.
Some people think this is a new concept and rant and rave and some who read slashdot whine and moan about it. Somehow the government and corporations are controlling people's minds. Read into your history a little bit. Around the turn of the last century there were dozens of newspapers in San Fransisco. It had grown so fast and was inhabited by so many different sorts of people that for a while it supported these several dozen newspapers. Then people began to homogenize and so did the different newspapers and publishing groups. Now you've got a handful of local newspapers in San Fransisco some with much larger circulations than others. See the correlation here? The web is going to be varied but there is also always going to be points where alot of people go to. Just because you've got a phone book with a million listings doesn't mean you're going to call them all, unless you're war dialing. Same goes for websites in directories.
Besides basic economics and social structures pervading the web researchers are often times not very well versed in the regions of the internet. Most research completely ignores IRC networks and message boards some of which are like slashdot and have nearly a bajillion people reading them per day. As well as IRC networks (which I know have declined a little bit in popularity) researchers seem to ignore instant messaging systems and their effect on the web. Alot of web users have abandoned e-mail lists, IRC networks, and message boards in lieu of instant messaging systems. I bet alot of people on analog modems probably IM more than they surf the web anymore. It doesn't require a whole lot of bandwidth and can be done on even old slow computers.
A good film image with good optics and lighting taken with good film will need about 20 megapixels in order to be comparable digitally. For shots most people would never notice the difference in unless blown up fairly large is from 6 to 9 megapixels, these numbers of course are assuming the colour range is higher than 8 bits. What camera makers are working on is the colour depth of digital cameras. Single CCD or CMOS cameras only have 8 bits of colour depth per pixel and in order to generate full colour RGB images interpolate the remaining 16 bits of colour information. Digital cameras also have problems with contrast since they're only getting 256 levels of it while negative film grabs about a thousand levels of contrast. Digital sensors also have blooming problems where bright pixels bleed over into neighboring pixels which prevents you from taking pictures with really fine contrast between pixels. As it stands colour film scanners are much better for high quality shots because they have adequate pixel resolution as well as colour depth to get as much information off the film as possible. When 5MP cameras get down to consumer quality that is when you can figure that digital camers really will replace celluloid film cameras. Crappy shots from disposable camers are about what you're getting out of the current line of 1-2MP cameras but for much higher prices. Right now the Canon D30 CMOS camera is one of the best you can buy but it costs several times more than my Rebel 2000 (and uses the same EOS lenses) and doesn't deliver the same quality. Though this argument enters a grey area when you compare 5x7 prints from a 35mm film camera and a 5x7 print from a high quality photo printer (unless you're talking terms of cost in which I still win:).
Did I mention mainframe systems? No I didn't. Of course those run Unix of some variant. I've never seen an ATM that ran anything more complex than DOS for the simple reason is multi-user time sharing systems are not right for the application. ATMs are NOT merely terminals to bigger systems. They are entirely dependant on mainframe systems to operate but the front end is not just a terminal emulator. Embedded systems often times don't have a traditional interface one might think of as a computer. A couple of buttons and LEDs are usually the only user interface embedded systems have. Most of the shit you listed DOES NOT run NT or Unix. Look it up man. Most systems just have some firmware which runs a piece of software on them while other have RTOSes. Both NT and any Unix variant are ill suited for these environments because they are time sharing systems. There is no fucking reason to have a multi-user OS running on something that only interfaces with other machines and never people. NT and most Unix variants you find have too many problems with latency and poorly manage high priority processing. With something like a computer driving or piloting some vehicle you need stuff to process in a set amount of time and handle errors seemlessly. You can't have an autopilot computer pop up an hourglass on screen saying its busy.
What planet are you living on? Unix runs on virtually no embeded systems or ATM machines or public terminals. You need to lay off the purple twinkies man. Suggesting sticking any Unix varient on these sorts of systems is ridiculous. Why in the holiest of holy fucks would you use a time sharing OS on an ATM machine or embedded computer? That is just ridiculous. Most embedded systems don't even run an OS, just some firmware that is definitely not related to Unix in any way. As for the comment about customized systems you're just tilting at windmills. Have you ever seen WinNT embedded? No you probably haven't because it doesn't look like Windows. Where do you come up with this stuff?
I wouldn't even classify that as thinking. How are you going to cache sites off the internet without a connection to it provided by somebody? Are you magically going to get one of your nodes added to a routing table somewhere? Man I can't wait to see whut my fucking naighbors who live next door to me and I could go to their fucking house and talk to them have to put on their website. I also bet that nobody will balk at paying for their own access equipment and they will all magically learn to use it because it is easier than subscribing to AOL.
That's absolutely ludicrous. Do you know how Microsoft makes so much money selling an OS without making the hardware? They don't fucking support it. When they sell an OEM copy of Windows to an OEM that company takes on the responsibility of supporting both Windows and the hardware they sell. If you buy a new Dell with Windows XP and something on it goes wrong you have to first call up Dell to get them to fix it, they may or may not consult Microsoft but you sure as hell don't call Microsoft first. They will forward your call to Dell or charge you 200$ for support fees. The only time Microsoft supports their OS is when you buy it directly from them which costs you a pretty penny.
Apple tried to license its OS to clone makers a couple years ago but ran into problems support wise because it ended up being that they were just subsidizing the manufacture of Mac systems to other manufacturers. This brought a temporary bounce in Mac market share but ultimately cost them money because they had to support all of the clones sold. So they tried increasing the licensing fees and the clone makers balked and Apple stopped licensing the OS. Software licensing is simple when you've got no stake in hardware.
If Apple didn't support the OS on OEM licenses (like Microsoft does) no one would buy the licenses because it would require tens of millions of dollars for them to build an entirely new support infrastructure for the new OS (which is also an argument against sticking Linux on their hardware). These OEMs would also be at the whim of hardware manufacturers that supported the new OS with drivers and support themselves. PC OEMs make their money because they can get components from just about anybody because everybody writes Windows drivers for their hardware. If Apple did support their x86 system they'd have to charge OEMs beaucoup cash or else they'd end up losing serious money having to support all the jackasses who couldn't understand the differences between Finder and Explorer. Why the fuck do you think Taligent and Rhapsody fucking fell through? Rhapsody (which started life as NeXT and eventually became OSX Server and so forth) was originally meant to be cross platform but would have meant Apple would definitely lose the hardware business. Apple is not a software company and I don't think they ever should be.
Slashdot Mirror
The problem with mp3.com was the artists had to sign over the rights to any songs mp3.com sold under the pay per play program. The same ass fucking took place just with a different entity behind the intense reaming. Maybe mp3.com acted better and didn't dick the artist around as much as one of the big four but the artist still had to fork over his or her rights to the work. For some I bet it was awesome to sell at least something without some uber-tough record contract squeezing out your creative spirit.
Performance of microkernels is definitely debatable. Microkernels on systems with slow context switching or a low number of active processes (like the PPC 603) perform rather badly because of a generally slow system and the required overhead of their design. However once you stick them on faster and more capable hardware you start to definitely see improcements. Monolithic kernels though run well in the environment where you need a little slimmer code because you're lacking speed/ability. Windows NT is still a microkernel design though just about everything runs in serverland as Win32 or userland as extensions of Win32. OSX is another good example of a microkernel design, one people seem to get confused. It is primarily a Mach kernel that runs a BSD 4.4 Lite based server and thus is compatible with BSD 4.4 Lite compatible software and functions. You can drop the BSD subsystem out of OSX and it will still work fine as the MacOS components based on Carbon and Cocoa don't require the BSD system at all.
Without the GNU project Linux as an OS would simply not exist. There MIGHT be a 1.x kernel floating around right now had there need no GNU project to actually do something with the kernel. Where do you think most of the shit redhat packages as their "base" system comes from? They sure as hell didn't write it themselves. It's also pretty retarded to say "why use HURD when there's Linux" because by the same logic you can say "why use Linux when there is [insert OS here]". How come Linux users are so quick to forget all of the stuff that had to happen for Linux to even exist. Shit man one wayward sperm and ther'd be no Linus to write Linux.
IANAKEOAS ( I am not a kernel expert of any sort) but microkernels differ from monolithic kernels is many ways other than hardware drivers running as services. They only pass messages between components meaning EVERYTHING is a server. Want to run Linux programs on a Mach kernel? Just run a Linux server and the software will run just fine. A good example of this is MkLinux and Windows NT. Windows NT runs a Win32 as a server but by the same token can run a POSIX server so POSIX compliant software also runs on it. MkLinux is a Mach kernel running a Linux server to run Linux software. Microkernels in theory are platform agnostic as they only pass messages between various servers. Most people except Linus Tovalds really dig microkernels and have put alot of work into them. Another advantage is you can remain platform agnostic in design yet run servers specific to hardware you're running on. You can go from running on a StrongARM system with 16 megs of RAM to a 32 processor x86 system using a majority of the same code. Notable microkernel based OSes include Windows NT, MacOS X, and of course HURD.
You're ignoring just about everything I said, that being big shared memory systems are better than clustered systems for particular jobs. If you want to serve half a gillion web pages or run a bunch of copies of FLUENT buy the Athlon cluster. Same with a big renderfarm doing batch rendering. If you want a realtime system you need a real computer. You said it yourself, you can malloc just about the entire system's memory for a single process. If say you need to run FLUENT with really big datasets clustering may or may not give you a price performance advantage. Instead of FLUENT I was thinking more along the lines of realtime video where an Onyx is feeding several channels of video each with specific content.
It seems to piss you off that somebody might suggest clustered systems are not the end all be all of computing. You seem to assume I know shit about Origin/Onyx systems when in fact I've read several essays written by SGI's engineers about them. I've also used older Origin systems (pre 3k systems) as well as lots of O2s doing realtime video work. I know the history of the XIO switching and the advent of the ccNUMA stuff. This doesn't make me an engineer but I'm also not talking out of my ass. I've seen plenty of specs for big and fancy Ethernet based Athlon clusters but they are dumb renderfarms. Its fine if you can break down some big problem into a bunch of easily digested packets for a cluster to munch on, in fact that's cool in my book if you can.
The P4 and Athlon beat out most other processors if compared CPU to CPU but after that they fail to scale. Adding N processors to a Ethernet based cluster does not scale well in terms of bandwidth and eventually you end up with diminishing returns and lots of wasted cycles because your cluster just doesn't feed its processors with enough data. It certainly won't process anything near realtime like a big iron system will. You should know this from experience and not ignore the failing of clusters in several cases.
Irix works well on the hardware it was written for. The original post claimed Irix was a POS and Linux could somehow magically be made for fill its role. I contend that Linux cannot and will not be made to fit this role because Irix has become too well adapted to its task. Trying to do the same with Linux would be damn near impossible if your goal was to make a Linux system that looks like a normal Linux system yet is running a 512 processor behemoth. I also didn't say anyone bought 32 processor machines to write code, they buy them to run code. Specifically code they've written. Shit man how many of your customers came to you saying "we've got this Fortran program we need to run in X amount of time, what do you suggest"? If the answer pisses YOU off then tough. Don't get your panties in a bunch because somebody didn't pee their pants when somebody said the word cluster.
I was just at the local mall earlier this afternoon. Whilst looking in Software Ect. I asked if I could play the XBox demo they had in the front window which had bee nturned off. The reason it was turned off is it had overheated a couple days before Christmas and they've yet to get a box to replace it with. Unless Rob lives in a universe parallel to my own with slightly different thermodynamics I don't think his comment was out of touch with reality.
Uh..with the ccNUMA structure of Origins a bad DIMM doesn't do much more than you're minus that many megs of memory overall. It is a switched architecture, the failure of one part of the system doesn't mean the whole thing is dead. If a processor module dies you just need to replace it. Have you seen the insides of C-Bricks? Unplug, unhook, slide out, slide in, hook, plug, done. You also make it seem like Irix is somehow a shitty OS on SGI hardware. Wow I want to spend a cool million of a huge Onyx and try to port Linux to it. Right. As for code being written for distributed machines that's just bullshit. Most if not everybody who's buying a system with more than 32 processors is writing their own software to run on it. Saying they are going to buy a MPP machine and partition it to run as a cluster is ridiculous. That is another point you are missing, Onyx and origin systems are NOT SMP systems, they are MPP (massively parallel processor) systems. SMP is where a couple processors populate a bus or have crossbar connects to one another. An individual C-Bick in an Origin might be considered an SMP system but the Origin itself is not. I'd really love to see you build a cluster than can do even half the things an Onyx system can. Renderfarms are just a bunch of retarded nodes that take individual frames and render them which is easily clustered. You don't buy an Onyx just to batch render a bunch of video frames. Your final point just doesn't make sense. When isn't an application based on the amount of memory and IO of a system? Who cares if an Athlon has a higher SPECfp score than a R12k. Clusters can't manipulate enormous datasets in realtime like MPP systems can.
Charging 3000$ for the CD set means that 99% of the jackasses who would use the GPL in order to buy something and then turn around and release it for free can't afford it while the 1% that can have to pay a pretty penny to be jackasses. I can pretty much assure you some jackass Linux zealot with no understanding on the GPL is sitting in his bedroom right now trying to figure out how he can raise 3k so he can be a folk hero by releasing the code an evil company is keeping secret. At the very least HP is giving some idiot something to do.
What the fuck are you talking about? Do you know anything about SGI hardware or software?
1. Saying SGI boxes do not scale completely invalidates any thought you've ever had in your life. Have you ever seen an Onyx/Origin system before? You add a brick and have twice the processing horse power or add 10 PCI (or any periphrial connection you ca nthink of) slots to do anything you want. You don't even have to turn the system off to add anything to it. Origins scale from 2 processor to 128 processor systems at the drop of a fucking hat. Using the Indy and O2 for your examples is just jackassery. Neither of which were designed for easy upgrades. Later revisions of the O2 when the Octane was released were easy as could be to upgrade. You turn the system off and slide out a processor modules and replace it with a new one. Voila.
2. Clusters fucking suck. Jesus get it through your Beowulf worshiping head that clusters are not the answer to all computation problems. LANL might build a bix cluster of PC systems to solve some embarrassingly redundant set of equations big fucking whoop. The big SGI systems are cache coherent and have messaging latencies measured in nanoseconds. A processor on one end of a 512 processor machine can talk to another processor with similar latency that a PC based system can talk to its own RAM. It is trivial to write software that spawns processes onto multiple nodes in a cluster, ccNUMA systems run as if their far flunt components were a single machine.
3. Linux cannot do that. If it COULD do that it would no longer be Linux and certainly would not be portable outside of SGI's hardware. Do you seriously think you can get good performance out of off the shelf software? Hell no. Optimizing software and good drivers doesn't mean anything. You've bitten far too deeply into Linux hype if you think it is even a contender to the grand daddy Unix systems.
Bringing up NewTek and ReelMagic cards is just fucking ridiculous. Wow you can have a hardware video processor that uses a PCI bus. That's rad...until you've seen an Origin in action running more than a dozen pipelines feeding 8 display channels. Try that with a PCI bus and see what happens. Nobody buys SGI for the logo.
A useful link would be here over at LinuxDevices.com. The board I might suggest to you in the Bitsy with its PCMCIA slot on board. With 16MB of RAm and 32MB of Flash ROM you'd have plenty of space to stick some software to run your car. Grab an off the shelf 802.11 PCMCIA card you can find drivers for and you've got it networked. The board also has USB which works for a camera and total of 21 digital IOs that you can use to control the components of the RC car. A bonus is the card also has audio in and out so you can turn your little car into a little talking robotic badass.
Strangely people talk about passive cooling in tower cases. Passive cooling would work much much better in a desktop chassis than in a tower. Reason being you could take a nice sized heat sink and stick it on your processor with plenty of headroom over it. With enough intake slits near the processor convetion would work pretty well for the most part. Heat sinks are perpendicular to the plane of the processor but making the plane of the processor perpendicular to the direction air is going to naturally flow ends up fucking up your cooling. More than a dozen models of Macs were designed this way and coupled with the fact the 603 disappated about 5 watts at the most made for a pretty quiet design overall, my PowerBook uses the same concept but instead of a heat sink it conducts heat into my lap. If you took a decent desktop chassis with a Tualatin P3, Celeron, or Via C3 you could probably get away with entirely passive cooling on the processor itself. As for the other noisy parts thats somebody else's problem.
Linux will never be like OSX because all of its zealot users are too anal. They clamour about choice like it's going out of style. With MacOS you write a program using one of a small number of APIs (Cocoa and Carbon specifically). If a program is written for OSX it just works. Linux folks want 20 different graphics toolkits. It is a natural tenent of the whole Bazaar concept that a good portion of all work overlaps in function. You're never going to see something like OSX come out of any completely open source project. The skill might be there but then to are the egos which make everybody go off and do their own thing.
What the fuck is this guy smoking? Like so many others have said this ought to be 10 pipe dreams for 2002. Not one thing he said made much sense at all and in general sounds like a half baked article he came up with at 6am to meet a 7am deadline.
1.) Purple twinky induced fantasy.
2.) Yet more fantasy, this time assuming users of Linux GUIs will be able to paste text between different applications written by different people. Back when there was growing dissatisfaction about IBM's licensing yet they are still the biggest computer company ever.
3.) This has been said for the past three years and has yet to happen. Why? Apps developed to scratch an itch are often not too broad in scope and have little intention of starting a paradigm. Apps intended to replace closed source counterparts rarely if ever achieve said couterpart's functionality. You end up with a system that doesn't talk to itself with a quarter of the features you could get for paying for something.
4.) People woken up from being cryogenically frozen for the past year could fucking tell you this.
5.) Right. Do you know what sort of systems REALLY classified data is stored on? Probably not. Three letter agencies don't exactly order their super secret computer systems from Dell with Windows 98 installed on them.
6.) No. That is just retarded. Unless several dozens apps somehow get ported to Linux magically this will not happen. No one is going to prepackage and OS they can't sell software for. Case in point BeOS.
7.) Maybe. Leo Laporte likes Linux but I've never heard him actually say anything he has really used it for. Wow you can replace a small handful of Windows programs with it that is sure to impress alot of people. A Linux TV show would be like a live action freshmeat.
8.) What the fuck? That is the most ridiculous thing in the entire article which in itself is notable. AOL on Linux would be like putting a vinyl interior in an Abrams tanks. It's asthetic vinyl...fitting to a fucking tank.
9.) *bong sound*
10.) see 9
It'd be damn near impossible to sell Linux to 99% of the people buying computers from retail sources. Even if you pushed it with millions of dollars in marketing you couldn't sell it to most people. If they couldn't call up a single number to get support if they have problems they won't buy the computer. It is that simple. No one wants to be refered to newsgroups and some crusty HOW-TOs when they run into a snag trying to get on the internet or when a program won't open because of some obscure library being in the wrong place. People are definitely not going to buy a system when they ask if they can edit some home movies or play video games and get laughed at by a sales person. Where's Linux's capability at? It doesn't matter if it can run for a year without rebooting. If there's nothing to use for a year for the retail computer buying public nobody's going to buy it.
Uh...the genetic difference between man and chimpanzee is about 1.6% which means we share 98.4% of our genetic material with chimpanzees. That 1.6% differences code all the phentotypes that makes us different from chimpanzees as well as our similarities.
When they were building the first PC they already had manufacturing rights to the 8088/8086 because they used the 8086 in the Displaywriter Intelligent Typewriter. Since they could already produce the chip it fit in well with their "off the shelf" component structure of the PC.
256 levels of contrast ratio is not nearly enough to match what the human eye can see. Colour negatives come far closer to what the eye can distinguish. Also there are only a very very small handful of cameras that offer more than 8 bits of quantization and these all cost more than 1000$ and then a small portion of those have interchangable lenses and the ability to do long term exposures. If you'd be so kind as to be specific as to why my price/performance assertion is bogus I'd really enjoy that. I'd love to see how a 300$ digital camera that takes pictures equivilent to a disposable Kodak camera at best beats even a crappy SLR camera. Maybe as you move up the rungs of camera quality high priced digital cameras might come out close to cost effective when compared to SLR camera but I really don't see it. I think I'd take my SLRs and my film scanned by professional scanning labs over a 5MP Canon until I scan upwards of 70 rolls a year which I don't happen to.
What about anything I said about bit depth is wrong? Besides I answered the question the guy asked and that was whether digital camers right now really compare to film which they don't. Price/performance wise they are completely uneconomical unless you have a point and shoot 35mm and take more than 40 rolls of film a year. People with those sorts of camers are lucky to develop 10 rolls a year. Here you you will be better informed read up. He says the exact same thing as me and is an admittedly better photographer. Fucking smug people.
Statistically all statistics lie in some form. Journalists statistically don't have any degrees besides journalism. Few journalists statistically know more about statistics than statisticians.
My point? You can infer anything from statistics, thats like the first chapter of a statistics textbook. So web eyeballs have narrowed their focus to a smaller number of websites in a given amount of time, big deal. If you go a little farther back in time you'll see the exact same thing as today, a majority of web users visited a handful of websites. Why? Lots of reasons. The biggest is only a smaller number of websites offered content the majority of web users even wanted to look at. Then there was a boom of websites that all offered the same thing packaged a little differently, some put blue bows on their piles of shit whilst others wrapped their shit in red bows. People liked the red bows more and thus now most of the blue bow sites are gone. Before the boom there were a handful of sites because no one thought much of the internet, now there are a handful of sites because people overvalued the internet.
Some people think this is a new concept and rant and rave and some who read slashdot whine and moan about it. Somehow the government and corporations are controlling people's minds. Read into your history a little bit. Around the turn of the last century there were dozens of newspapers in San Fransisco. It had grown so fast and was inhabited by so many different sorts of people that for a while it supported these several dozen newspapers. Then people began to homogenize and so did the different newspapers and publishing groups. Now you've got a handful of local newspapers in San Fransisco some with much larger circulations than others. See the correlation here? The web is going to be varied but there is also always going to be points where alot of people go to. Just because you've got a phone book with a million listings doesn't mean you're going to call them all, unless you're war dialing. Same goes for websites in directories.
Besides basic economics and social structures pervading the web researchers are often times not very well versed in the regions of the internet. Most research completely ignores IRC networks and message boards some of which are like slashdot and have nearly a bajillion people reading them per day. As well as IRC networks (which I know have declined a little bit in popularity) researchers seem to ignore instant messaging systems and their effect on the web. Alot of web users have abandoned e-mail lists, IRC networks, and message boards in lieu of instant messaging systems. I bet alot of people on analog modems probably IM more than they surf the web anymore. It doesn't require a whole lot of bandwidth and can be done on even old slow computers.
A good film image with good optics and lighting taken with good film will need about 20 megapixels in order to be comparable digitally. For shots most people would never notice the difference in unless blown up fairly large is from 6 to 9 megapixels, these numbers of course are assuming the colour range is higher than 8 bits. What camera makers are working on is the colour depth of digital cameras. Single CCD or CMOS cameras only have 8 bits of colour depth per pixel and in order to generate full colour RGB images interpolate the remaining 16 bits of colour information. Digital cameras also have problems with contrast since they're only getting 256 levels of it while negative film grabs about a thousand levels of contrast. Digital sensors also have blooming problems where bright pixels bleed over into neighboring pixels which prevents you from taking pictures with really fine contrast between pixels. As it stands colour film scanners are much better for high quality shots because they have adequate pixel resolution as well as colour depth to get as much information off the film as possible. When 5MP cameras get down to consumer quality that is when you can figure that digital camers really will replace celluloid film cameras. Crappy shots from disposable camers are about what you're getting out of the current line of 1-2MP cameras but for much higher prices. Right now the Canon D30 CMOS camera is one of the best you can buy but it costs several times more than my Rebel 2000 (and uses the same EOS lenses) and doesn't deliver the same quality. Though this argument enters a grey area when you compare 5x7 prints from a 35mm film camera and a 5x7 print from a high quality photo printer (unless you're talking terms of cost in which I still win :).
Did I mention mainframe systems? No I didn't. Of course those run Unix of some variant. I've never seen an ATM that ran anything more complex than DOS for the simple reason is multi-user time sharing systems are not right for the application. ATMs are NOT merely terminals to bigger systems. They are entirely dependant on mainframe systems to operate but the front end is not just a terminal emulator. Embedded systems often times don't have a traditional interface one might think of as a computer. A couple of buttons and LEDs are usually the only user interface embedded systems have. Most of the shit you listed DOES NOT run NT or Unix. Look it up man. Most systems just have some firmware which runs a piece of software on them while other have RTOSes. Both NT and any Unix variant are ill suited for these environments because they are time sharing systems. There is no fucking reason to have a multi-user OS running on something that only interfaces with other machines and never people. NT and most Unix variants you find have too many problems with latency and poorly manage high priority processing. With something like a computer driving or piloting some vehicle you need stuff to process in a set amount of time and handle errors seemlessly. You can't have an autopilot computer pop up an hourglass on screen saying its busy.
What planet are you living on? Unix runs on virtually no embeded systems or ATM machines or public terminals. You need to lay off the purple twinkies man. Suggesting sticking any Unix varient on these sorts of systems is ridiculous. Why in the holiest of holy fucks would you use a time sharing OS on an ATM machine or embedded computer? That is just ridiculous. Most embedded systems don't even run an OS, just some firmware that is definitely not related to Unix in any way. As for the comment about customized systems you're just tilting at windmills. Have you ever seen WinNT embedded? No you probably haven't because it doesn't look like Windows. Where do you come up with this stuff?
I wouldn't even classify that as thinking. How are you going to cache sites off the internet without a connection to it provided by somebody? Are you magically going to get one of your nodes added to a routing table somewhere? Man I can't wait to see whut my fucking naighbors who live next door to me and I could go to their fucking house and talk to them have to put on their website. I also bet that nobody will balk at paying for their own access equipment and they will all magically learn to use it because it is easier than subscribing to AOL.
A WLAN connected to what exactly?
That's absolutely ludicrous. Do you know how Microsoft makes so much money selling an OS without making the hardware? They don't fucking support it. When they sell an OEM copy of Windows to an OEM that company takes on the responsibility of supporting both Windows and the hardware they sell. If you buy a new Dell with Windows XP and something on it goes wrong you have to first call up Dell to get them to fix it, they may or may not consult Microsoft but you sure as hell don't call Microsoft first. They will forward your call to Dell or charge you 200$ for support fees. The only time Microsoft supports their OS is when you buy it directly from them which costs you a pretty penny.
Apple tried to license its OS to clone makers a couple years ago but ran into problems support wise because it ended up being that they were just subsidizing the manufacture of Mac systems to other manufacturers. This brought a temporary bounce in Mac market share but ultimately cost them money because they had to support all of the clones sold. So they tried increasing the licensing fees and the clone makers balked and Apple stopped licensing the OS. Software licensing is simple when you've got no stake in hardware.
If Apple didn't support the OS on OEM licenses (like Microsoft does) no one would buy the licenses because it would require tens of millions of dollars for them to build an entirely new support infrastructure for the new OS (which is also an argument against sticking Linux on their hardware). These OEMs would also be at the whim of hardware manufacturers that supported the new OS with drivers and support themselves. PC OEMs make their money because they can get components from just about anybody because everybody writes Windows drivers for their hardware. If Apple did support their x86 system they'd have to charge OEMs beaucoup cash or else they'd end up losing serious money having to support all the jackasses who couldn't understand the differences between Finder and Explorer. Why the fuck do you think Taligent and Rhapsody fucking fell through? Rhapsody (which started life as NeXT and eventually became OSX Server and so forth) was originally meant to be cross platform but would have meant Apple would definitely lose the hardware business. Apple is not a software company and I don't think they ever should be.