There definitely was a "sonic drill" article a few months back. It had something to do with the particular shape of the focusing chamber (I don't remember the shape, was it bell-shaped?) defeating the formation of shockwaves. These were explained as the primary loss of energy during the creation of a high energy directed sound stream... It might have been in Popular Science. Still, this is supposed to be much better in terms of control.
Someone should cut down this guy's ritlin doseage. Seriously though, I never thought I'd see a representative of a company so blatantly admit that it's really the consumer they're fighting, especially without choosing J. Random Artist to be their poster child. At least he's honest, I suppose. If you can call that honest.
Personally, I'll still stick with mp3.com. I personally find quite a bit to like there, esp. the comedy sections... Gotta love that "Internet Help Desk" skit:-)
No kidding... I was able to pull over 500KB/sec downstream over RoadRunner in upstate New York, and about 50-60KB upstream. I was never bothered about the web server I had running, though a few of their techs knew it existed, and the only port they blocked was sendmail (and I believe that's a good thing, with all the spammers out there...) I'd go back to RR in a second if it was of the same quality out here.
PacBell DSL has really disappointed me, I top out around 150KB downstream and 13.5KB upstream. I'm pretty close to the telecom too. The bandwidth and service limitations that @home imposes in the Silicon Valley area make cable a non-starter, even worse than DSL. I suppose everything, even bandwidth, is more expensive out here though:-)
It has less to do with them deciding on a standard, and more to do with a standard being forced upon them through vendor lock-in. Until Linux has everything necessary for them to do their work efficiently (it's getting close, folks) some people have to use it. As much as I hate to admit it, we're not there yet.
I'm honestly interested, if you've tried this... How well does GLX work over two discrete video cards in a dual headed config? The Xi Graphics people had a demo at LinuxWorld showing a G400 dual head doing 3D accel spanning both screens flawlessly. Loki had Flight Gear running on what the demonstrator called a dual headed G400 as well, though I believe it was just mirroring at the time for the projector. Pretty nifty stuff that didn't look too bad at all...
To be honest, considering such 'artists' as Britney Spears, NSync and the Backstreet Boys, I'd have to say that the subsidized music industry _is_ full of complete crap. The industry as it stands pushes a small number of artists, ones that it's either sure will make money, or those that can be molded into something that will. These 'artists' are the ones that are likely to receive any reasonable amount of money from their music in the short term, and frankly they probably would never have gotten anywhere without those companies. If that's the sort of thing you want to hear, the current system works.
Now consider this. If you go over to MP3.com, you can listen to whatever songs are posted. These songs come from artists who are able to put together their music on their own, without a record company behind them, and obviously it didn't take them $500,000 to put it together. If you think something is crap, you simply move on to another artist. If you like something enough that you'd want it, you can always purchase the CD.
How much was that again? You mean you can get an album for less than $10, and the artist gets more per CD than they would have from a record company? And horror of horrors, there's noone preventing you from listening to the music, for free, as much as you like before the purchase?
It's obvious that there's a serious paradigm shift happening in the record industry. As has always happened, they'll either adapt or die. If they disappear, I for one won't miss them.
Think Babelfish will come up with a translator for all this? 1337-speak and Prince jargon?:-)
Seriously though, while I can't say I've always been the biggest fan of TAFKA Prince's music, I agree dead on with his philosophy. That a person would imply that music *shouldn't* be free is pretty disgusting. To use Prince's analogy (sort of...) If you love your music, you'll want it to be free, such that as many people can experience it as possible. If you love money more than your music, well... I suppose that shows what sort of artist you are.
You know, there's just something wrong when you have to change your name to regain your freedom. Some might call that terrorism. I call it BS:-)
I suggest any who haven't yet done so, check out the about:mozilla which is at least in the M17 and nightly M18 builds... We have an entry from a heretoforth unrevealed chapter/edition:-)
Aside from that, I found the Linux/x86 M18 build tonight to be wonderful, better than the M17 build by quite a margin. The interface is getting much faster, though it's not quite as fast as native GTK. If this level of improvement continues, we're in for a fast browser, and soon:-)
I seem to recall that there were announcements quite a bit before the first pre-betas were released. I could certainly be wrong about this, but various hints of it started popping up quite a bit before there was anything available to download. Back when WindowMaker was undergoing development I believe... It's been quite a while:-) Hardly seems possible.
Regardless, the KDE project has had more time to work on the desktop "problem", and it's not surprising that they have come up with the better solution at this time.
Such a project exists. But the KDE project itself is currently more interested in making their environment better. You're more than welcome to join the Harmony project though. I'm actually of the opinion that the GNOME folks would have done better to work on this than compete with KDE for little other reason, it seems childish. But part of what makes life so interesting is that one person's opinion is not the only valid opinion. I wish luck to both KDE and GNOME.
Miguel also said that UNIX sucks. More likely it's his approach at the project which is sub-optimal, for instance an object oriented language lends itself to writing object oriented environments. Procedural languages, well, don't.
I'm amazed that GNOME is as far along as it is, and while there's more to be done, we should remember it takes more effort to do this sort of thing in C than C++. (Remember too that the KDE project was around for at least a year before _it_ was released.)
Don't get me wrong, I like KDE. I use it daily, and think it has the best chance to "win the desktop" for 'NIX if anything on X will. They've done a lot of things right, and created a wonderful environment which has allowed me to quickly assist many people getting into UNIX for the first time.
That doesn't mean that GNOME isn't headed in the right direction. They just need time. I look forward to seeing what they have in store for the coming year.
Remember that KDE started over a year before GNOME was even thought of. They have a bit of catching up to do. Frankly have done an admirable job considering they had to develop an object model on top of C, them's the breaks. Don't be surprised that KDE is about a year ahead of GNOME in many areas, software development takes time.
I'm curious, what keyboard shortcuts exactly do you find lacking? I use KDE as my primary desktop, and I seldom touch the mouse unless I'm using Netscape. Of course I usually use alt-f2 to launch programs and open up file windows by typing their location (the tab completion is quite nice,) but it seems quite keyboard-friendly to this anti-rodent KDE user.
There is a shellscript that comes with BASH (/usr/doc/bash-doc-2.04/adventure.sh in a Mandrake 7.1 install) that could be used as a users' shell. It's no more advanced than the parser which came with the x00 point adventure games, but it's interesting as a simple proof of concept.
Oddly enough, it doesn't seem to want to execute now:-) I remember playing with it a few years ago, it's rather humorous.
Sounds pretty nifty, I have to wonder what the memory architecture will look like in those beasts. Think if I drop by SGI they'll let me play with it?:-) (j/k)
I wonder if SGI will start putting those Cray-Link interfaces on the SGI/Linux boxen:-) Now THAT would make a cluster... Not likely in any case, but we can dream.
>Bullshit. One doesn't follow the other in any >local sense.
Nice language there. In any case, they actually follow quite closely. Never forget that such things have happened in the past, books _have_ been banned for their content and teachers for that matter _have_ been prohibited in some places from teaching sexual education, for example. There are many corporations/organizations which would find it convenient to have this kind of control over what you think... Remember that many a war has been fought over rights you might believe to be silly. (Tea taxes?!)
Consider the current lawsuits from Mattel and the DVD consortium against, esentially, the world. The freedom to disseminate information is being raped from people in both of these cases. I'm sure you'd be happy to see the Bible banned due to its violent and sexual content, cultural/religious significance notwithstanding. If it's a small enough freedom, of no importance to yourself, you probably won't miss it. Not until you're part of the next fringe group they wish to target. Or are you one of those religious zealot types?
>Typical whine #8... you may not like what I'm >doing, but if you don't support it they will take >away your freedom to! >Right.
Well, I ask you this: Just what is it that makes you free now? I submit that you're probably less free now than you have ever been in your life, and it is only going to get worse if current trends contine. When you're asking yourself "How did this happen to me?" remember this day, and your indifference. I'm sure it will be of great comfort to you to know that most people don't care. Absurd? We'll see.
We can hope it will work, but what guarantee do we really have? I'm not hedging my bets on a closed source "solution." Closed source, and especially non-standard drivers where an established standard exists (how many GLIDE-only 3D apps do we see these days?) have been too much trouble in the past. I don't intend to wait for promised support again.
When we can be absolutely sure that software linked with their libraries will work with Mesa, and vice versa, then we're headed in the right direction. Even then, how can we be sure that they'll give us a driver for the next C library version? The next XF86 version? Will they support legacy binary versions for older, binary-only programs? If so, how soon will these be released, considering that they will probably have newer cards to support and Linux will be taking the back seat to Windows in their labs? What assurance do we have that if NVidia goes out of business we'll still be able to use our video cards with newer versions of Linux? Will we be able to use these cards with FreeBSD/other OSs if we decide to switch our software platform? What about hardware, and non-x86 chips? Obviously we won't be able to fix the bugs ourselves, so are we still dealing with unsupported drivers which will likely have several glitches? Who do we turn to when they fail if we can't fix it ourselves?
Until they're tested, we have only promises. On some of these points we don't even have that. With NVidia's history in the promise department, and especially with their 'NDA' excuses, I'm not buying into it. Hardware drivers are _no_ place for closed source, not if we want to be able to use that hardware in the future.
"A $275 paperweight? Can I get back to you on that?"
I can understand what you're saying to a point, but this is a very short term view of the problem. NVidia has decided that they will play by different rules, which will only bring incomptatibilities into the development of Linux 3D software. We can hope that most companies will ensure that the NVidia drivers work with app X, and that this app also supports standard DRI. If we don't have the source, we have no guarantee that they will even support the next version of X. Why do they feel they have to make it difficult on their customers? These same customers don't want their work/applications to be wasted next year when they need to change their software to use a newer video card, or vice versa.
As for the legal encumbrances, I consider this to be a bug, as surely as a segmentation fault. They could release a set of specifications to allow these parts of the driver to be filled in. If they only release binaries, it means that it isn't going to work on all platforms *unless* they decide to support Alpha, PowerPC, etc. If they decide to ignore these ports, they will not be able to support the entire market whereas hardware with an open source driver would. When you are looking at high end 3D, the Alpha platform in particular should come to mind. Customers with these computers must wait for NVidia's decision to support this platform before even considering their cards.
You have raised an interesting point by bringing up people who don't care about open source principles. If they are using Linux to do their 3D work, they are likely to seek the services of someone who is familiar with the market. This is entirely reasonable as they will likely want some support. At this point, the hardware choice may well be in the hands of someone who knows a bit more about the hows and whys of said support under Linux. Under the current situation, this person is more likely to choose, say, the Matrox card. If they feel they can ignore the open source proponents simply because they(we!) are a small segment of the market, they should take into account that we often make purchasing decisions for others.
Again, I understand your position on this. I definitely appreciate the work you have put into the GLX drivers, and can understand why you feel the closed source drivers might be a good option. But as much as I enjoy a good game of Quake III, I won't run it on a card which might not work with the next C library version. Until Linux has a stable ABI, and perhaps even then, there is no excuse for any company which cares about its customers to provide binary only drivers for their hardware. I won't support such a company, and actively encourage others to seek alternatives.
There are a few reasons why the Solaris model is better in theory, but one really stands out.
Specifically, the fact that you can map N threads to N processes does help. Rather than mapping 1 thread to 1 process, as Linux currently does, mapping say 5 threads to 1 process can be very helpful. When you intend to have a large amount of threads, very common on a server (a threaded web server for example, perhaps running java servlets) you have the potential to fill the process table quickly. If you can map N threads to each process, you are N times less likely to fill the table, and thus can scale higher.
More simply put, if you're limited to about 65000 threads, and open about 16000 connections with 3 threads each, there's no way to launch another process/thread without killing an existing one. On a machine with 8 processors and 4 GB RAM, you're probably wasting a lot of your processing power simply because of this. More users cannot in theory connect, regardless of your resources. If you can map multiple threads to each process, then the problem is solved (or at least postponed.) More users can connect up to the theoretical limits of the hardware/process table.
It may not be faster in all situations to do this, but it's IMHO the Right Way(tm). If Linux is to truly challenge the high end server OS', this type of functionality will be necessary at some point. Hopefully this is the answer you were looking for, albeit a bit late:-)
I'm running Mandrake 7.0 (2nd ISO bugfix release) on a K6/2-450, and it seems to be working quite well. There were a few small issues with the compiler being unhappy (I got rid of that by installing egcs RPMs, re-compiling GCC 2.95 and installing that) but otherwise it's been quite well mannered.
For reference, my system includes 128MB RAM, a Matrox G100, 2 ethernet cards (1 realtek ISA/1 3C905B PCI), an ISA aha152x SCSI card with a zip drive attached, a Voodoo 2 8MB, 3 IDE HDs, a SB16 non-pnp sound card and an IDE CD-ROM drive. Quite an assortment of hardware, all autodetected, and it runs very smoothly (unlike Windows 2000 Pro, which doesn't like my 3com ethernet card...)
I'd chance to guess you might be having a problem with the CPU overheating. This is very common with the higher speed K6/2 and K6/3 chips. Memory is also a possible problem, there are many odd errors with RAM that only rear their ugly heads when certain access patterns are used, these are possibly exascerbated by Mandrake's "Pentium Optimizations."
If you have multiple DIMMs/SIMM banks, try removing some of the RAM to see if that makes a difference. If it works properly, put the RAM you just removed in the box by itself and try it again. If it fails, you've likely just found your problem.
You can also try turning down the voltage on your processor by 1/10th of a volt, your processor will probably run stably at this voltage. This was recommended by Tyan on their web site, it won't damage your processor to try it and it works quite well on my 1590S. Dropped the temp by quite a few degrees on my CPU, and lower temps mean fewer failures.
MPEG compression is computation-heavy and lossy, not exactly what you'd want to run your word processor or photo editing software through. We'd need something a bit more lightweight to push the huge volume of data in real time. Perhaps RLE encoding would do the trick, it doesn't get much lighter than that and would likely work wonders on the (*usually) uncompressed data streams sent to the video card. Using this on top of a pure delta protocol would probably get something like what you're looking for with minimal encoding cost.
There's no reason that this couldn't work for FMV data though, barring costs and complexity it could be the logical extension of hardware-assisted MPEG cards. Sounds very yummy... It'd have to be programmable/upgradeable though, to keep up with new technologies. Who wants to have to purchase a new 36-inch display because a better video compression algorithm came out?
> As far as a SCSI interface - SCSI may be fast > for storage, but it gets absolutely torched by > even PCI.
To my knowledge, the 32 bit PCI bus (running at 33MHz) has a throughput of around 132MB/sec. The latest Ultra SCSI-3 specification, at 160MB/sec, can't be fully taken advantage of in this case. Of course, 64 bit PCI is a different story, as is AGP1x, where in both cases the bandwidth is doubled by the bus width getting you 264MB/sec. (AGP 2x and 4x each double the bandwidth further by increasing the clock speed.)
Either way, unless you're updating the entire screen at this resolution, you don't need this type of bandwidth on the bus just to display video. You need only pass the information necessary to modify the display, and let the video card's acceleration engine (either 2D or 3D) handle the rest. AGP even gets you the benny of offloading the image data from the PCI bus with sideband addressing, giving the rest of the system room to breathe next to that hungry video card.
Since the video card's engine is doing the actual manipulation of the DAC, that's the part that needs to be this fast. With today's 250MHz and (easily) faster RAMDACs, this isn't quite the problem it might seem at first. Unless you are trying to do full motion video at that resolution (good luck getting that bandwidth!) you'll probably do fine with AGP2x or better. Now, if the VGA connector starts running into problems transferring that amount of data, you could have a real problem on your hands... But until that time, I'd say current high end videocards could handle this class of display.
Slashdot Mirror
There definitely was a "sonic drill" article a few months back. It had something to do with the particular shape of the focusing chamber (I don't remember the shape, was it bell-shaped?) defeating the formation of shockwaves. These were explained as the primary loss of energy during the creation of a high energy directed sound stream... It might have been in Popular Science. Still, this is supposed to be much better in terms of control.
Someone should cut down this guy's ritlin doseage. Seriously though, I never thought I'd see a representative of a company so blatantly admit that it's really the consumer they're fighting, especially without choosing J. Random Artist to be their poster child. At least he's honest, I suppose. If you can call that honest.
:-)
Personally, I'll still stick with mp3.com. I personally find quite a bit to like there, esp. the comedy sections... Gotta love that "Internet Help Desk" skit
No kidding... I was able to pull over 500KB/sec downstream over RoadRunner in upstate New York, and about 50-60KB upstream. I was never bothered about the web server I had running, though a few of their techs knew it existed, and the only port they blocked was sendmail (and I believe that's a good thing, with all the spammers out there...) I'd go back to RR in a second if it was of the same quality out here.
:-)
PacBell DSL has really disappointed me, I top out around 150KB downstream and 13.5KB upstream. I'm pretty close to the telecom too. The bandwidth and service limitations that @home imposes in the Silicon Valley area make cable a non-starter, even worse than DSL. I suppose everything, even bandwidth, is more expensive out here though
Most users I deal with simply ask this question:
What else is there?
It has less to do with them deciding on a standard, and more to do with a standard being forced upon them through vendor lock-in. Until Linux has everything necessary for them to do their work efficiently (it's getting close, folks) some people have to use it. As much as I hate to admit it, we're not there yet.
Then again, we have only just begun.
I'm honestly interested, if you've tried this... How well does GLX work over two discrete video cards in a dual headed config? The Xi Graphics people had a demo at LinuxWorld showing a G400 dual head doing 3D accel spanning both screens flawlessly. Loki had Flight Gear running on what the demonstrator called a dual headed G400 as well, though I believe it was just mirroring at the time for the projector. Pretty nifty stuff that didn't look too bad at all...
And I suppose El Camino Virtual must be Castro/Moffet. (I actually live just off this road, and work at Moffet Field...)
Virtual Mr. Hankey
There's no need, 101 is almost always 'slashdotted' :-)
To be honest, considering such 'artists' as Britney Spears, NSync and the Backstreet Boys, I'd have to say that the subsidized music industry _is_ full of complete crap. The industry as it stands pushes a small number of artists, ones that it's either sure will make money, or those that can be molded into something that will. These 'artists' are the ones that are likely to receive any reasonable amount of money from their music in the short term, and frankly they probably would never have gotten anywhere without those companies. If that's the sort of thing you want to hear, the current system works.
Now consider this. If you go over to MP3.com, you can listen to whatever songs are posted. These songs come from artists who are able to put together their music on their own, without a record company behind them, and obviously it didn't take them $500,000 to put it together. If you think something is crap, you simply move on to another artist. If you like something enough that you'd want it, you can always purchase the CD.
How much was that again? You mean you can get an album for less than $10, and the artist gets more per CD than they would have from a record company? And horror of horrors, there's noone preventing you from listening to the music, for free, as much as you like before the purchase?
It's obvious that there's a serious paradigm shift happening in the record industry. As has always happened, they'll either adapt or die. If they disappear, I for one won't miss them.
Mr. Hankey
Think Babelfish will come up with a translator for all this? 1337-speak and Prince jargon? :-)
:-)
Seriously though, while I can't say I've always been the biggest fan of TAFKA Prince's music, I agree dead on with his philosophy. That a person would imply that music *shouldn't* be free is pretty disgusting. To use Prince's analogy (sort of...) If you love your music, you'll want it to be free, such that as many people can experience it as possible. If you love money more than your music, well... I suppose that shows what sort of artist you are.
You know, there's just something wrong when you have to change your name to regain your freedom. Some might call that terrorism. I call it BS
Mr. Hankey
I suggest any who haven't yet done so, check out the about:mozilla which is at least in the M17 and nightly M18 builds... We have an entry from a heretoforth unrevealed chapter/edition :-)
:-)
Aside from that, I found the Linux/x86 M18 build tonight to be wonderful, better than the M17 build by quite a margin. The interface is getting much faster, though it's not quite as fast as native GTK. If this level of improvement continues, we're in for a fast browser, and soon
I seem to recall that there were announcements quite a bit before the first pre-betas were released. I could certainly be wrong about this, but various hints of it started popping up quite a bit before there was anything available to download. Back when WindowMaker was undergoing development I believe... It's been quite a while :-) Hardly seems possible.
Regardless, the KDE project has had more time to work on the desktop "problem", and it's not surprising that they have come up with the better solution at this time.
http://www.gnu.org/software/harmony/harmony.html
Such a project exists. But the KDE project itself is currently more interested in making their environment better. You're more than welcome to join the Harmony project though. I'm actually of the opinion that the GNOME folks would have done better to work on this than compete with KDE for little other reason, it seems childish. But part of what makes life so interesting is that one person's opinion is not the only valid opinion. I wish luck to both KDE and GNOME.
Miguel also said that UNIX sucks. More likely it's his approach at the project which is sub-optimal, for instance an object oriented language lends itself to writing object oriented environments. Procedural languages, well, don't.
I'm amazed that GNOME is as far along as it is, and while there's more to be done, we should remember it takes more effort to do this sort of thing in C than C++. (Remember too that the KDE project was around for at least a year before _it_ was released.)
Don't get me wrong, I like KDE. I use it daily, and think it has the best chance to "win the desktop" for 'NIX if anything on X will. They've done a lot of things right, and created a wonderful environment which has allowed me to quickly assist many people getting into UNIX for the first time.
That doesn't mean that GNOME isn't headed in the right direction. They just need time. I look forward to seeing what they have in store for the coming year.
For now, isn't having a choice wonderful?
Remember that KDE started over a year before GNOME was even thought of. They have a bit of catching up to do. Frankly have done an admirable job considering they had to develop an object model on top of C, them's the breaks. Don't be surprised that KDE is about a year ahead of GNOME in many areas, software development takes time.
I'm curious, what keyboard shortcuts exactly do you find lacking? I use KDE as my primary desktop, and I seldom touch the mouse unless I'm using Netscape. Of course I usually use alt-f2 to launch programs and open up file windows by typing their location (the tab completion is quite nice,) but it seems quite keyboard-friendly to this anti-rodent KDE user.
Howdy ho!
:-) I remember playing with it a few years ago, it's rather humorous.
There is a shellscript that comes with BASH (/usr/doc/bash-doc-2.04/adventure.sh in a Mandrake 7.1 install) that could be used as a users' shell. It's no more advanced than the parser which came with the x00 point adventure games, but it's interesting as a simple proof of concept.
Oddly enough, it doesn't seem to want to execute now
Mr. Hankey
Sounds pretty nifty, I have to wonder what the memory architecture will look like in those beasts. Think if I drop by SGI they'll let me play with it? :-) (j/k)
I wonder if SGI will start putting those Cray-Link interfaces on the SGI/Linux boxen :-) Now THAT would make a cluster... Not likely in any case, but we can dream.
>Bullshit. One doesn't follow the other in any
>local sense.
Nice language there. In any case, they actually follow quite closely. Never forget that such things have happened in the past, books _have_ been banned for their content and teachers for that matter _have_ been prohibited in some places from teaching sexual education, for example. There are many corporations/organizations which would find it convenient to have this kind of control over what you think... Remember that many a war has been fought over rights you might believe to be silly. (Tea taxes?!)
Consider the current lawsuits from Mattel and the DVD consortium against, esentially, the world. The freedom to disseminate information is being raped from people in both of these cases. I'm sure you'd be happy to see the Bible banned due to its violent and sexual content, cultural/religious significance notwithstanding. If it's a small enough freedom, of no importance to yourself, you probably won't miss it. Not until you're part of the next fringe group they wish to target. Or are you one of those religious zealot types?
>Typical whine #8... you may not like what I'm
>doing, but if you don't support it they will take
>away your freedom to!
>Right.
Well, I ask you this: Just what is it that makes you free now? I submit that you're probably less free now than you have ever been in your life, and it is only going to get worse if current trends contine. When you're asking yourself "How did this happen to me?" remember this day, and your indifference. I'm sure it will be of great comfort to you to know that most people don't care. Absurd? We'll see.
We can hope it will work, but what guarantee do we really have? I'm not hedging my bets on a closed source "solution." Closed source, and especially non-standard drivers where an established standard exists (how many GLIDE-only 3D apps do we see these days?) have been too much trouble in the past. I don't intend to wait for promised support again.
When we can be absolutely sure that software linked with their libraries will work with Mesa, and vice versa, then we're headed in the right direction. Even then, how can we be sure that they'll give us a driver for the next C library version? The next XF86 version? Will they support legacy binary versions for older, binary-only programs? If so, how soon will these be released, considering that they will probably have newer cards to support and Linux will be taking the back seat to Windows in their labs? What assurance do we have that if NVidia goes out of business we'll still be able to use our video cards with newer versions of Linux? Will we be able to use these cards with FreeBSD/other OSs if we decide to switch our software platform? What about hardware, and non-x86 chips? Obviously we won't be able to fix the bugs ourselves, so are we still dealing with unsupported drivers which will likely have several glitches? Who do we turn to when they fail if we can't fix it ourselves?
Until they're tested, we have only promises. On some of these points we don't even have that. With NVidia's history in the promise department, and especially with their 'NDA' excuses, I'm not buying into it. Hardware drivers are _no_ place for closed source, not if we want to be able to use that hardware in the future.
"A $275 paperweight? Can I get back to you on that?"
I can understand what you're saying to a point, but this is a very short term view of the problem. NVidia has decided that they will play by different rules, which will only bring incomptatibilities into the development of Linux 3D software. We can hope that most companies will ensure that the NVidia drivers work with app X, and that this app also supports standard DRI. If we don't have the source, we have no guarantee that they will even support the next version of X. Why do they feel they have to make it difficult on their customers? These same customers don't want their work/applications to be wasted next year when they need to change their software to use a newer video card, or vice versa.
As for the legal encumbrances, I consider this to be a bug, as surely as a segmentation fault. They could release a set of specifications to allow these parts of the driver to be filled in. If they only release binaries, it means that it isn't going to work on all platforms *unless* they decide to support Alpha, PowerPC, etc. If they decide to ignore these ports, they will not be able to support the entire market whereas hardware with an open source driver would. When you are looking at high end 3D, the Alpha platform in particular should come to mind. Customers with these computers must wait for NVidia's decision to support this platform before even considering their cards.
You have raised an interesting point by bringing up people who don't care about open source principles. If they are using Linux to do their 3D work, they are likely to seek the services of someone who is familiar with the market. This is entirely reasonable as they will likely want some support. At this point, the hardware choice may well be in the hands of someone who knows a bit more about the hows and whys of said support under Linux. Under the current situation, this person is more likely to choose, say, the Matrox card. If they feel they can ignore the open source proponents simply because they(we!) are a small segment of the market, they should take into account that we often make purchasing decisions for others.
Again, I understand your position on this. I definitely appreciate the work you have put into the GLX drivers, and can understand why you feel the closed source drivers might be a good option. But as much as I enjoy a good game of Quake III, I won't run it on a card which might not work with the next C library version. Until Linux has a stable ABI, and perhaps even then, there is no excuse for any company which cares about its customers to provide binary only drivers for their hardware. I won't support such a company, and actively encourage others to seek alternatives.
There are a few reasons why the Solaris model is better in theory, but one really stands out.
:-)
Specifically, the fact that you can map N threads to N processes does help. Rather than mapping 1 thread to 1 process, as Linux currently does, mapping say 5 threads to 1 process can be very helpful. When you intend to have a large amount of threads, very common on a server (a threaded web server for example, perhaps running java servlets) you have the potential to fill the process table quickly. If you can map N threads to each process, you are N times less likely to fill the table, and thus can scale higher.
More simply put, if you're limited to about 65000 threads, and open about 16000 connections with 3 threads each, there's no way to launch another process/thread without killing an existing one. On a machine with 8 processors and 4 GB RAM, you're probably wasting a lot of your processing power simply because of this. More users cannot in theory connect, regardless of your resources. If you can map multiple threads to each process, then the problem is solved (or at least postponed.) More users can connect up to the theoretical limits of the hardware/process table.
It may not be faster in all situations to do this, but it's IMHO the Right Way(tm). If Linux is to truly challenge the high end server OS', this type of functionality will be necessary at some point. Hopefully this is the answer you were looking for, albeit a bit late
Mr. Hankey
Howdyho!
I'm running Mandrake 7.0 (2nd ISO bugfix release) on a K6/2-450, and it seems to be working quite well. There were a few small issues with the compiler being unhappy (I got rid of that by installing egcs RPMs, re-compiling GCC 2.95 and installing that) but otherwise it's been quite well mannered.
For reference, my system includes 128MB RAM, a Matrox G100, 2 ethernet cards (1 realtek ISA/1 3C905B PCI), an ISA aha152x SCSI card with a zip drive attached, a Voodoo 2 8MB, 3 IDE HDs, a SB16 non-pnp sound card and an IDE CD-ROM drive. Quite an assortment of hardware, all autodetected, and it runs very smoothly (unlike Windows 2000 Pro, which doesn't like my 3com ethernet card...)
I'd chance to guess you might be having a problem with the CPU overheating. This is very common with the higher speed K6/2 and K6/3 chips. Memory is also a possible problem, there are many odd errors with RAM that only rear their ugly heads when certain access patterns are used, these are possibly exascerbated by Mandrake's "Pentium Optimizations."
If you have multiple DIMMs/SIMM banks, try removing some of the RAM to see if that makes a difference. If it works properly, put the RAM you just removed in the box by itself and try it again. If it fails, you've likely just found your problem.
You can also try turning down the voltage on your processor by 1/10th of a volt, your processor will probably run stably at this voltage. This was recommended by Tyan on their web site, it won't damage your processor to try it and it works quite well on my 1590S. Dropped the temp by quite a few degrees on my CPU, and lower temps mean fewer failures.
Hope this helps!
Mr. Hankey
MPEG compression is computation-heavy and lossy, not exactly what you'd want to run your word processor or photo editing software through. We'd need something a bit more lightweight to push the huge volume of data in real time. Perhaps RLE encoding would do the trick, it doesn't get much lighter than that and would likely work wonders on the (*usually) uncompressed data streams sent to the video card. Using this on top of a pure delta protocol would probably get something like what you're looking for with minimal encoding cost.
There's no reason that this couldn't work for FMV data though, barring costs and complexity it could be the logical extension of hardware-assisted MPEG cards. Sounds very yummy... It'd have to be programmable/upgradeable though, to keep up with new technologies. Who wants to have to purchase a new 36-inch display because a better video compression algorithm came out?
*Texture compression notwithstanding.
> As far as a SCSI interface - SCSI may be fast
> for storage, but it gets absolutely torched by
> even PCI.
To my knowledge, the 32 bit PCI bus (running at 33MHz) has a throughput of around 132MB/sec. The latest Ultra SCSI-3 specification, at 160MB/sec, can't be fully taken advantage of in this case. Of course, 64 bit PCI is a different story, as is AGP1x, where in both cases the bandwidth is doubled by the bus width getting you 264MB/sec. (AGP 2x and 4x each double the bandwidth further by increasing the clock speed.)
Either way, unless you're updating the entire screen at this resolution, you don't need this type of bandwidth on the bus just to display video. You need only pass the information necessary to modify the display, and let the video card's acceleration engine (either 2D or 3D) handle the rest. AGP even gets you the benny of offloading the image data from the PCI bus with sideband addressing, giving the rest of the system room to breathe next to that hungry video card.
Since the video card's engine is doing the actual manipulation of the DAC, that's the part that needs to be this fast. With today's 250MHz and (easily) faster RAMDACs, this isn't quite the problem it might seem at first. Unless you are trying to do full motion video at that resolution (good luck getting that bandwidth!) you'll probably do fine with AGP2x or better. Now, if the VGA connector starts running into problems transferring that amount of data, you could have a real problem on your hands... But until that time, I'd say current high end videocards could handle this class of display.