Because CPUs are like cars. They are designed to go much faster than the speed limits, although with speed comes greater risk. Some people like to race on the freeway at midnight (dumb bastards), while some like to crank their CPU beyond the posted speed limit (which is much less dangerous:)
There are a significant number of changes to the kernel internals, especially w/ regard to the VFS layer (and internal layer that lets the kernel treat each filesystem uniformly) Making all these changes causes many things to break, until they are all updated to conform to the new internal interface layer. The VFS is expected to evolve even more for 2.5/2.6.
Why make all these changes, instead of minor incremental changes to the internals? So that the kernel can handle such things as advanced new filesystems (XFS, ReiserFS, etc.) w/ larger file sizes, advanced features, etc. Also, Linux has incorporated a logical volume manager (LVM) allowing people to link filesystems, grow filesystems, etc. Without updating the VFS abstraction layer, XFS and others would have to resort to kludges, that would be difficult to maintain in the future.
Basically, Linux's VFS layer is a VERY powerful concept, allowing one to plug new file systems into the kernel, and have the rest of the kernel automatically use it without tons of rewriting. However, as people get more ambitious with their filesystem dreams, the VFS must evolve to provide the proper abstraction of features.
Also, one more neat thing (unrelated to VFS). I'm running 2.3-pre6 (released today), and the Soundblaster Live! OSS drivers have been included (emu10k1). That is great news! Considering I never expected my SBLive to work under Linux when I bought, it is great to see that it is now in the mainstream releases.
However, as a current TNT2 owner, I will NOT upgrade to another card that doesn't have an open source driver available (or at least possible). So over the summer, I will be changing from Nvidia to a Matrox, 3dfx, or even (uggh) ATI (this after I switched AWAY from ATI due to crappy windows drivers...)
I'm glad NVidia has released something, but I'm upset I didn't just buy the Matrox G400 6 months ago.
Yeah, but there are other unpatented "algorithms" that would identically translate the uncompressed version, without themselves being considered decompression algorithms. I think if you make a compressor (or decompressor) that does NOT build an LZW tree and use it for searching and replacing the current message, you are not violating the patent. In other words, handling the "uncompressed" GIFs, without building a tree of some sort, could not be construed (in any reasonable interpretation) as infringing. (IANAL)
I think it would be entirely possible to build a compatible compressor/decompressor that did not use any of Welch's patented enhancements to LZ78 in order to create and decode GIFs. Welch's improvements seem to mainly deal with efficiency (time & memory), but a simpler LZ78 algorithm could probably be made compatible (with some tricks). Of course, that probably would still attract a lawsuit, and wouldn't be as good as just changing to PNGs. I'll have to look into it...
Just a correction; everywhere you said LZH (Lempel-Ziv-Huffman), change that to LZW (Lempel-Ziv-Welch).
The first is free of any patents (as far as I know) and is based on an LZ77 algorithm, with the added step of Huffman coding for the sliding window pointer values. The second is the patented method (where the patent pertains to the methods for dynamically building, rescaling, and otherwise fiddling with a tree of previous symbol strings) and is based on an LZ78 algorithm.
Re:The problem with Rambus compared to SDRAM...
on
Will Rambus Go Bust?
·
· Score: 2
However, one must consider Intel's (and Rambus's) long-term strategy. In the future, density will continue to increase, allowing greater amounts of memory per module (mitigating the RDRAM latency issue somewhat), and the idea is that boosting frequency down the line (for RDRAMs) will be easier than adding more lines to SDRAM, or trying to boost SDRAM frequency (due to its more parallel nature). This is the same argument that all newer high speed serial protocols are using (FireWire, USB, future SCSI), based on the idea that serial protocols can be clocked MUCH higher than parallel.
Now, for memory, this doesn't help us today (or even for the next couple of years probably). But five or ten years down the line, a method like RAMBUS's might prove to have been the correct long term choice.
However, playing Devil's Advocate, I'd also have to believe that the SDRAM type technologies are going to push into higher and higher clock rates as well (due mainly to the relatively short paths that the signals must travel), and give RDRAM no clear performance win for quite a while. Also, it isn't clear to me that RDRAM is the right implementation of an idea, even if the idea is a good one.
But the main point is that while RAMBUS doesn't have a clear performance advantage now, it may in 5 years or so. But I wouldn't want to bet the farm on it.
Do GIF's have a mode whereby they use RLE (run-length encoding) for compression, rather than LZW, and if so, why not just use that mode? Given that, for the web, most GIF's are used as icons and are pretty small, compression shouldn't be a huge issue these days (and fancier, photoquality type things can just use JPEG...).
Someone could take a sampling of Web GIFs and compare their size w/ RLE vs. LZW, and see how much extra space would be used. I may be wrong about GIFs being able to use RLE; someone correct me if that's the case. If so, one could see how much LZW saves on representative GIFs that are uncompressed (and uncompressed GIFs have *got* to be possible:)
Check out this link about the Controversy over Problem F, and look for Dominic_Mazzoni's post (+5 interesting). I'll quote from that article:
Take a look at problem F. (Here are the contest problems in PDF if you're interested.) In a nutshell, you're given a complete directed graph, and you need to return the average length of all shortest paths between all pairs of nodes. The problem explicitly stated that you will only be given graphs in which there exists a path from every node to every other.
This is not a hard problem to work out, but anyone who has had a formal course in computer science ought to recognize that the Floyd-Warshall all-pairs shortest path algorithm is designed to solve exactly this problem. Then all you have to do is add up all of the elements of the matrix and divide by n * (n-1).
Except that the judges made a mistake, and tested our input using a graph that was not connected - in other words, there were nodes that could not reach other nodes via a directed path. This would not be a big deal, except that the problem explicitly stated that this would not occur. (Input validation is never a part of this contest.) Furthermore, without further explanation it is unclear how these nonexistent paths should affect the average. It turns out that the judges' solution was not counting these paths, and averaging only the paths that existed. Some teams did this by accident, and others (including Waterloo) figured it out only after submitting multiple runs and incurring large penalties. My team was one of the many that did not figure out the judges' mistake, so we did not get credit for the problem, even though our solution was certainly correct as the problem was worded. If we had received credit we would have had four problems correct, possibly putting us in the top ten. Of course, if we had received credit right away, we might not have wasted so much time figuring out what was wrong with our solution and we could have solved another problem in that time. Of course, many other teams were in a similar situation, so I have no idea what the final ranking would have been, but clearly it would have been different.
But I really have to wonder about the Panda. They may be cute and all, but if mother nature really wanted (a lot of) them around, why did she give them such a mindlessly defective reproductive system?
Well, given their environment, it may make sense. They developed in fairly remote areas of China, and probably had few natural predators. Furthermore, their diet (mostly bamboo, which is low in nutrients, and relatively slow to replace compared to other vegetation) could not support a high profundity, which would probably have led to overgrazing and starvation. They evolved to maintain a reasonably low population through slow reproduction (since predation wasn't helping).
Now, with human encroachment and loss of habitat, their reproductive habits have become a liability (only offset by their "cuteness", which now helps the species to remain viable)
Boy, I must have missed that memo. Let's see do corportations:
Take taxes from you? Yes, they obtain government subsidies through blackmail (ie. "We will move our operations overseas if you don't subsidize us."), and can use those subsidies in ways that do not contribute back to the local or national economy.
Start wars? Yes, there are numerous turn of the century insurrections that came about as a result of corporations and cartels manipulating the governments (including the U.S.) to use force. Our recent invation of Panama can largely be viewed as an attempt to retain control of the canal, keep shipping fees reasonable, etc.
Engage in gross acts of waste? Good lord yes! Wanton pollution, overfishing, overharvesting of resources, are all examples of waste. The "corporate welfare", as I mentioned before, can also be considered as a waste of government funds, when it is used by corporations for means that are non-beneficial for those who payed them (ie. the taxpayers)
Hmm... guess not... I'll take corporations over government any day of the week. You're in luck, corporations are increasingly able to use government as a tool to obtain their own ends, by proxy. You may prefer corporations to government, but then, you really have no choice, do you?
It amazes me that many Intel chips now come with upwards of 512k (or more) of cache memory, which is the total amount memory that the original 8086 PC could support. You could run an old Lotus 123 program completely in the processor cache of a modern chip (including data and a bunch of TSRs). This really puts the infamous B. Gates quote, about no one needing more than 640k, into perspective.
AFAIK it is proven that any higher-times DES give no significant security bonus over TripleDES regardless of the greater keylength.
A quick search of my literature reveals no such proof, although pg.366 of Applied Cryptography implies that Quintuple DES offers improved resistance to meet-in-the-middle attacks, as it certainly does for exhaustive search.
Michigan is PARTICULARLY evil in this respect. There are numerous such cases, including a famous one where a young man was put in jail for LIFE for providing information on where to obtain a controlled substance, without ever being in possesion of it, or profitting from its sale, etc. As far as I know, he is still in jail for it.
I encourage anyone with a link to the details to follow up this post w/ relevant info.
Unless you are able to find a huge (for large values of "huge") amount of computing power, your chance of finding the key, as an individual, is zero (plus epsilon, where epsilon is smaller than a Slashdot troller's vocabulary). However, joining a team means you may actually have a chance of winning SOMETHING, even if it is just accolades, or the right to vote on which charity gets the cash.
Anyone who is cracking RC5 for the money, is deluding themselves. Do it for the sex appeal!
You are right to point out that TripleDES is still the most tested and proven cipher in the field, despite DES being much maligned these days. It should be noted that TripleDES is thought of so highly simply because it has been around for so long. Blowfish, IDEA, and others are too new to be completely trusted, although they may very well be as secure or more secure than TripleDES.
Unfortunately, there is generally no way of formally proving security for these block ciphers, other than showing that it can resist various known real and hypothetical attacks (opening the possibility that some currently unknown technique could be found that crack a block cipher wide open)
So, in other words, DES and TripleDES have stood the test of time, but there is no guarantee that TripleDES is more secure than a newer block cipher. Personally, I'd use QuintupleDES if I were really ultra paranoid, otherwise I like Blowfish because of its speed. (And it is free for anyone to use and implement)
Actually, this just validates ECC, as it was fully expected that the 108-bit challenge was feasible (ie. could be broken in reasonable time with massive computing power). The linked site seems to indicate it took about 5 months (searching about 76% of the problem space), which is roughly what the challenge issuers anticipated.
The 97-bit challenge was broken (by the same project) in a couple of months, also in line w/ prediction.
The next challenge is 131 bits, which should take a LOT more time and power to break. The challengers can use these results to justify their claim that the 161+ bit challenges are infeasible.
FWIW, people in the field believe a 160 bit ECC problem to be roughly equivalent to 1024 bits of RSA, which is currently WAY beyond what can be feasibly attacked (barring no advances in factoring technology)
ECC has a bright future and these challenges only serve to emphasize that point.
It is the simple fact that an armed populace is the best defense against tyranny.
I would argue that an informed and concerned population is the best defense against tyranny. Giving everyone in the U.S. a gun won't stop the subtle and persistent erosion of our rights by corporate-backed government stooges.
At last summer's LinuxWorld Expo, Linus commented on khttpd, saying that (among other things) it was such a well written module, and an instructive example of how to write Linux driver code, that he thought it should be included just on that basis. The driver's scope is simple enough, that it should be very stable, and not cause problems elsewhere in the kernel.
Those who complain about Microsoft's delays, and shout about "vaporware", are just a vocal minority, and don't represent the community as a whole. I don't care about Microsoft delaying Windows 2000 (even though I have to make strategic decisions based on Windows releases), and I don't mind 2.4 being later than promised (since I can always use the development kernel and patches, most of which are stable enough for my needs).
So, no, we are NOT all hypocrites, though, undoubtedly, there are a few among us.
Holy shit, that is NOT true! "Cordwainer Bird" is a derogatory pseudonym used by Harlan Ellison, whose highest praise for Gene Roddenberry these days is that "he was a lying sack of shit!" Don't know how Roddenberry got his "nickname", but "Cordwainer Bird" has nothing to do with it.
VMWare can host each machine in less than 128 Megs, and doesn't use much memory itself. So if you can configure the guest OS to use say 32 megs, then each VMWare instance takes roughly 32megs + some overhead (lets say 8 megs for the sake of argument).
Yes, you may want a machine with lots of RAM to run multiple OS's at once, but that doesn't mean that each instance will use up lots of RAM. I have a machine w/ 256 Meg RAM, and host multiple NT guests running under Linux, w/ no problem.
Slashdot Mirror
Because CPUs are like cars. They are designed to go much faster than the speed limits, although with speed comes greater risk. Some people like to race on the freeway at midnight (dumb bastards), while some like to crank their CPU beyond the posted speed limit (which is much less dangerous :)
There are a significant number of changes to the kernel internals, especially w/ regard to the VFS layer (and internal layer that lets the kernel treat each filesystem uniformly) Making all these changes causes many things to break, until they are all updated to conform to the new internal interface layer. The VFS is expected to evolve even more for 2.5/2.6.
Why make all these changes, instead of minor incremental changes to the internals? So that the kernel can handle such things as advanced new filesystems (XFS, ReiserFS, etc.) w/ larger file sizes, advanced features, etc. Also, Linux has incorporated a logical volume manager (LVM) allowing people to link filesystems, grow filesystems, etc. Without updating the VFS abstraction layer, XFS and others would have to resort to kludges, that would be difficult to maintain in the future.
Basically, Linux's VFS layer is a VERY powerful concept, allowing one to plug new file systems into the kernel, and have the rest of the kernel automatically use it without tons of rewriting. However, as people get more ambitious with their filesystem dreams, the VFS must evolve to provide the proper abstraction of features.
Also, one more neat thing (unrelated to VFS). I'm running 2.3-pre6 (released today), and the Soundblaster Live! OSS drivers have been included (emu10k1). That is great news! Considering I never expected my SBLive to work under Linux when I bought, it is great to see that it is now in the mainstream releases.
However, as a current TNT2 owner, I will NOT upgrade to another card that doesn't have an open source driver available (or at least possible). So over the summer, I will be changing from Nvidia to a Matrox, 3dfx, or even (uggh) ATI (this after I switched AWAY from ATI due to crappy windows drivers...)
I'm glad NVidia has released something, but I'm upset I didn't just buy the Matrox G400 6 months ago.
Yeah, but there are other unpatented "algorithms" that would identically translate the uncompressed version, without themselves being considered decompression algorithms. I think if you make a compressor (or decompressor) that does NOT build an LZW tree and use it for searching and replacing the current message, you are not violating the patent. In other words, handling the "uncompressed" GIFs, without building a tree of some sort, could not be construed (in any reasonable interpretation) as infringing. (IANAL)
I think it would be entirely possible to build a compatible compressor/decompressor that did not use any of Welch's patented enhancements to LZ78 in order to create and decode GIFs. Welch's improvements seem to mainly deal with efficiency (time & memory), but a simpler LZ78 algorithm could probably be made compatible (with some tricks). Of course, that probably would still attract a lawsuit, and wouldn't be as good as just changing to PNGs. I'll have to look into it...
Just a correction; everywhere you said LZH (Lempel-Ziv-Huffman), change that to LZW (Lempel-Ziv-Welch).
The first is free of any patents (as far as I know) and is based on an LZ77 algorithm, with the added step of Huffman coding for the sliding window pointer values. The second is the patented method (where the patent pertains to the methods for dynamically building, rescaling, and otherwise fiddling with a tree of previous symbol strings) and is based on an LZ78 algorithm.
However, one must consider Intel's (and Rambus's) long-term strategy. In the future, density will continue to increase, allowing greater amounts of memory per module (mitigating the RDRAM latency issue somewhat), and the idea is that boosting frequency down the line (for RDRAMs) will be easier than adding more lines to SDRAM, or trying to boost SDRAM frequency (due to its more parallel nature). This is the same argument that all newer high speed serial protocols are using (FireWire, USB, future SCSI), based on the idea that serial protocols can be clocked MUCH higher than parallel.
Now, for memory, this doesn't help us today (or even for the next couple of years probably). But five or ten years down the line, a method like RAMBUS's might prove to have been the correct long term choice.
However, playing Devil's Advocate, I'd also have to believe that the SDRAM type technologies are going to push into higher and higher clock rates as well (due mainly to the relatively short paths that the signals must travel), and give RDRAM no clear performance win for quite a while. Also, it isn't clear to me that RDRAM is the right implementation of an idea, even if the idea is a good one.
But the main point is that while RAMBUS doesn't have a clear performance advantage now, it may in 5 years or so. But I wouldn't want to bet the farm on it.
Excuse me if this was a joke and I missed it, but if you are serious, then your logic makes me want to get a lobotomy.
Do GIF's have a mode whereby they use RLE (run-length encoding) for compression, rather than LZW, and if so, why not just use that mode? Given that, for the web, most GIF's are used as icons and are pretty small, compression shouldn't be a huge issue these days (and fancier, photoquality type things can just use JPEG...).
:)
Someone could take a sampling of Web GIFs and compare their size w/ RLE vs. LZW, and see how much extra space would be used. I may be wrong about GIFs being able to use RLE; someone correct me if that's the case. If so, one could see how much LZW saves on representative GIFs that are uncompressed (and uncompressed GIFs have *got* to be possible
Readers of Alan Moore's (and Dave Gibbon's :) Watchmen know all about the smiley face on Mars.
Check out this link about the Controversy over Problem F, and look for Dominic_Mazzoni's post (+5 interesting). I'll quote from that article:
Take a look at problem F. (Here are the contest problems in PDF if you're interested.) In a nutshell, you're given a complete directed graph, and you need to return the average length of all shortest paths between all pairs of nodes. The problem explicitly stated that you will only be given graphs in which there exists a path from every node to every other.
This is not a hard problem to work out, but anyone who has had a formal course in computer science ought to recognize that the Floyd-Warshall all-pairs shortest path algorithm is designed to solve exactly this problem. Then all you have to do is add up all of the elements of the matrix and divide by n * (n-1).
Except that the judges made a mistake, and tested our input using a graph that was not connected - in other words, there were nodes that could not reach other nodes via a directed path. This would not be a big deal, except that the problem explicitly stated that this would not occur. (Input validation is never a part of this contest.) Furthermore, without further explanation it is unclear how these nonexistent paths should affect the average. It turns out that the judges' solution was not counting these paths, and averaging only the paths that existed. Some teams did this by accident, and others (including Waterloo) figured it out only after submitting multiple runs and incurring large penalties. My team was one of the many that did not figure out the judges' mistake, so we did not get credit for the problem, even though our solution was certainly correct as the problem was worded. If we had received credit we would have had four problems correct, possibly putting us in the top ten. Of course, if we had received credit right away, we might not have wasted so much time figuring out what was wrong with our solution and we could have solved another problem in that time. Of course, many other teams were in a similar situation, so I have no idea what the final ranking would have been, but clearly it would have been different.
Because it had logo that whooshed into frame over the spinning Earth. This would have been all wrong.
But I really have to wonder about the Panda. They may be cute and all, but if mother nature really wanted (a lot of) them around, why did she give them such a mindlessly defective reproductive system?
Well, given their environment, it may make sense. They developed in fairly remote areas of China, and probably had few natural predators. Furthermore, their diet (mostly bamboo, which is low in nutrients, and relatively slow to replace compared to other vegetation) could not support a high profundity, which would probably have led to overgrazing and starvation. They evolved to maintain a reasonably low population through slow reproduction (since predation wasn't helping).
Now, with human encroachment and loss of habitat, their reproductive habits have become a liability (only offset by their "cuteness", which now helps the species to remain viable)
And I remember that it cost a boatload of money to fix it, like say (in Dr. Evil voice) one MILLLLION dollars...
Actually, I don't remember the exact cost, but CGI rendering was MUCH more expensive back then.
Boy, I must have missed that memo. Let's see do corportations:
Take taxes from you?
Yes, they obtain government subsidies through blackmail (ie. "We will move our operations overseas if you don't subsidize us."), and can use those subsidies in ways that do not contribute back to the local or national economy.
Start wars?
Yes, there are numerous turn of the century insurrections that came about as a result of corporations and cartels manipulating the governments (including the U.S.) to use force. Our recent invation of Panama can largely be viewed as an attempt to retain control of the canal, keep shipping fees reasonable, etc.
Engage in gross acts of waste?
Good lord yes! Wanton pollution, overfishing, overharvesting of resources, are all examples of waste. The "corporate welfare", as I mentioned before, can also be considered as a waste of government funds, when it is used by corporations for means that are non-beneficial for those who payed them (ie. the taxpayers)
Hmm... guess not... I'll take corporations over government any day of the week.
You're in luck, corporations are increasingly able to use government as a tool to obtain their own ends, by proxy. You may prefer corporations to government, but then, you really have no choice, do you?
It amazes me that many Intel chips now come with upwards of 512k (or more) of cache memory, which is the total amount memory that the original 8086 PC could support. You could run an old Lotus 123 program completely in the processor cache of a modern chip (including data and a bunch of TSRs). This really puts the infamous B. Gates quote, about no one needing more than 640k, into perspective.
AFAIK it is proven that any higher-times DES give no significant security bonus over TripleDES regardless of the greater keylength.
A quick search of my literature reveals no such proof, although pg.366 of Applied Cryptography implies that Quintuple DES offers improved resistance to meet-in-the-middle attacks, as it certainly does for exhaustive search.
Michigan is PARTICULARLY evil in this respect. There are numerous such cases, including a famous one where a young man was put in jail for LIFE for providing information on where to obtain a controlled substance, without ever being in possesion of it, or profitting from its sale, etc. As far as I know, he is still in jail for it.
I encourage anyone with a link to the details to follow up this post w/ relevant info.
Unless you are able to find a huge (for large values of "huge") amount of computing power, your chance of finding the key, as an individual, is zero (plus epsilon, where epsilon is smaller than a Slashdot troller's vocabulary). However, joining a team means you may actually have a chance of winning SOMETHING, even if it is just accolades, or the right to vote on which charity gets the cash.
Anyone who is cracking RC5 for the money, is deluding themselves. Do it for the sex appeal!
You are right to point out that TripleDES is still the most tested and proven cipher in the field, despite DES being much maligned these days. It should be noted that TripleDES is thought of so highly simply because it has been around for so long. Blowfish, IDEA, and others are too new to be completely trusted, although they may very well be as secure or more secure than TripleDES.
Unfortunately, there is generally no way of formally proving security for these block ciphers, other than showing that it can resist various known real and hypothetical attacks (opening the possibility that some currently unknown technique could be found that crack a block cipher wide open)
So, in other words, DES and TripleDES have stood the test of time, but there is no guarantee that TripleDES is more secure than a newer block cipher. Personally, I'd use QuintupleDES if I were really ultra paranoid, otherwise I like Blowfish because of its speed. (And it is free for anyone to use and implement)
Actually, this just validates ECC, as it was fully expected that the 108-bit challenge was feasible (ie. could be broken in reasonable time with massive computing power). The linked site seems to indicate it took about 5 months (searching about 76% of the problem space), which is roughly what the challenge issuers anticipated.
The 97-bit challenge was broken (by the same project) in a couple of months, also in line w/ prediction.
The next challenge is 131 bits, which should take a LOT more time and power to break. The challengers can use these results to justify their claim that the 161+ bit challenges are infeasible.
FWIW, people in the field believe a 160 bit ECC problem to be roughly equivalent to 1024 bits of RSA, which is currently WAY beyond what can be feasibly attacked (barring no advances in factoring technology)
ECC has a bright future and these challenges only serve to emphasize that point.
See the link on the Certicom ECC challenge for more info.
It is the simple fact that an armed populace is the best defense against tyranny.
I would argue that an informed and concerned population is the best defense against tyranny. Giving everyone in the U.S. a gun won't stop the subtle and persistent erosion of our rights by corporate-backed government stooges.
At last summer's LinuxWorld Expo, Linus commented on khttpd, saying that (among other things) it was such a well written module, and an instructive example of how to write Linux driver code, that he thought it should be included just on that basis. The driver's scope is simple enough, that it should be very stable, and not cause problems elsewhere in the kernel.
Those who complain about Microsoft's delays, and shout about "vaporware", are just a vocal minority, and don't represent the community as a whole. I don't care about Microsoft delaying Windows 2000 (even though I have to make strategic decisions based on Windows releases), and I don't mind 2.4 being later than promised (since I can always use the development kernel and patches, most of which are stable enough for my needs).
So, no, we are NOT all hypocrites, though, undoubtedly, there are a few among us.
Holy shit, that is NOT true! "Cordwainer Bird" is a derogatory pseudonym used by Harlan Ellison, whose highest praise for Gene Roddenberry these days is that "he was a lying sack of shit!" Don't know how Roddenberry got his "nickname", but "Cordwainer Bird" has nothing to do with it.
VMWare can host each machine in less than 128 Megs, and doesn't use much memory itself. So if you can configure the guest OS to use say 32 megs, then each VMWare instance takes roughly 32megs + some overhead (lets say 8 megs for the sake of argument).
Yes, you may want a machine with lots of RAM to run multiple OS's at once, but that doesn't mean that each instance will use up lots of RAM. I have a machine w/ 256 Meg RAM, and host multiple NT guests running under Linux, w/ no problem.