Number 1 seems a bit wrong. The Via Padlock engine does two things - secure PRNG and it has a Montgomery multiplier. The secure PRNG doesn't help as that's not what the spy process is measuring. The multiplier is called either once, or twice, in the exponentation loop depending on a bit in the key. The leakage channel is the branch that decides whether or not to call the second multiplication. Is the Via Padlock that quick that it prevents the branch choice from leaking though the branch table?
I went to VON (Voice On the Net) 2000 where a lot of people were already exhibiting working systems that implement what the patent describes. The patent is a dead duck. Tons of prior art, and an entire tradeshow devoted to the fact that it was obvious to 'one proficient in the field'. SIP was already around then - designed to interface between different implementations of software phones.
You have to remember that lots of shit patents are filed, but then never used. This will sit in a cupboard forever as it is not worth the money to try and defend it.
This is a false argument. Even if you find two things which are not dependant on each other, it does not follow that there are no dependancies in the entire game. It's pretty damn obvious that the video and sound output DO have a sequential dependency on the controller input.
No the argument is correct. Although there are lots of sequential dependencies in the game, there is also lots of code that can be parallelised. The parallelisation is limited by the length of the longest sequential chain - this is another phrasing of Ahmdahl's law that the OP quoted. Everything within a timeslice (say a frame) will have a dependency on the IO, and the final output of the frame and sound will be dependent on everything in the timeslice, but this is not a large chain. One frame is typically ouput while the next is being computed (pipelining) - so the two tasks can be performed in parallel if you accept a one-frame latency. This is what most games do when they double-buffer the graphics. So then you are only concerned with the dependencies *within* a timeslice. Is the physics code, the graphics code, the AI code or the sound code dependent on each other within a timeslice? No. Read the article on Steam / Halflife Episode 2 sometime, it explains this coarse-grain parallelism well.
Well actually they do, or at least you have to assume they do until you've checked otherwise. Or are the bullets passing through the enimies with no effect? Sounds like a pretty silly game to me.
Now you've slipped back to fine-grain parallelism. Here's a clue: do I need to check the bullet against *every* object in the game? Wouldn't that be a bit silly? How about nearby ones only? Oooh, this would need some kind of spatial data-structure like an oct-tree. Then lets split the timestep into two-phases, on the first phase we'll compute all interactions between objects that are well within their spatial sub-division. We'll then broadcast these changes, and then process the areas around intersections. Gosh, a sequetial dependency chain of two steps - that's a constant by the way. And potential parallelism for as many processors as we have sub-divisions (or buckets if you will) in the spatial structure. This is why physics code can be parallelised well.
The comment about electrostatics applies because if you are considering non-local fields of interactions then this approach is borked and you need to consider n-squared interactions. But even then you can simulate in discrete timesteps and trade latency for potential parallelism.
In summary - there are no inter-module sequential dependencies because of double buffering (this is called ping-ponging in the GPGPU community). The intra-module sequential dependencies can be minimised by grouping spatial regions into a single calculation, although as Valve have pointed out this is difficult. And in fact there is 30 years of research into how to solve this problem. For a working example have a look at GPGPU where there is a paper showing a million particle simulation working in parallel on a GPU.
Yeah that's one way to do it. If you can assume the filesystem hasn't been compromised then you could store the executable on disk, otherwise build it at runtime.
RSA isn't the problem. The implementation of RSA on a modern processor is the problem. Moving to another algorithm wouldn't guarantee a lack of side channels. One way around this would be to specialise the algorithm with your own private key. This would unroll all of the loops, and decide the branches statically. If you assume that the machine is not compromised, then this executable could be stored as read-only for your account. If the machine is compromised enough for a non-priviledged process to read your private data then you don't need SBPA - you're toast.
Which should be plainly obvious. You're going to be playing 1 game, not 4 games with no interrelation to each other.
Really? So if I pick the sound code and the graphics code as an example you're telling me that there will be a sequential dependency between them? Just because two things are related does not mean that they are dependent.
For example, with physics you'll be able to parallelize the physics of ALL THE OBJECTS THAT DON'T INTERACT WITH EACH OTHER.
Ah, using capitals to reinforce your point. Let me correct your error:
ALL THE OBJECTS THAT DON'T directly INTERACT WITH EACH OTHER.
I'm sure you can tell which word I've added. Now think about this question - do you think that physics simulations on graphics hardware, or the PhysX board are sequential in nature? Although direct interaction implies communication, and hence a sequential dependency in the code, we live in a (mostly) Euclidean universe, and that limits the number of direct interactions. If you're simulating force interactions like electostatics between particles then everything has a direct interaction. But in a game the simulation is of collisions, and gravity. In this case not all objects directly affect each other. Hence they can be clustered into separate groups which can be processed in parallel. Going back to the cell as the example, I'd expect there are more than 8 such groups in every game simulation.
You are assuming that the parallel code and the sequential code have some sort of dependency that force them to run in sequence, otherwise the speedup is 4x for the Cell, rather than 2.9x. Obviously as the 25% bottleneck is reduced, the speedup is greater. The majority of code in a game is parallelisable - although it is not always easy to write it that way. The 25% is more like 2.5% for sequencing IO. All of the graphics code can be parallelised (assuming multiple viewport you can just do software SLI). AI is heavily parallelisable (especially the crappy rule-based search that current games use). Physics simulation is a classic parallel processing problem.
I would expect to see near complete utilisation of the Cell, right around the time that Sony introduces the PS4...
Today. I was really suprised to see a leaflet in the supermarket today for a 100% green energy supplier who will price-match any standard tariff. This means that they're not the cheapest on the market, but they are as cheap as most people pay. This is a uk company, so I don't know if there is something similar for the majority of the us slashdot readership.
For Brits who are interested in 100% green energy at market prices: Ecotricity
Q: Does it make sense for a Photoshop plugin to be used.
A: NO! As I said, a procedural texture tool would have nothing to do with bitmaps.
After your long-winded argument with the other two posters I think that you've browbeat them into submission. But you are still wrong. I don't think anyone in this thread thought that procedural textures had anything to do with bitmaps, you've brought this up as a strawman. The real question, as you well know, is what kind of tool would an artist like in order to generate procedural textures?
Here's a clue. It's not a source-code editor. I've done work in both procedural texuring, and procedural geometry and I use a text editor. But I'm a programmer and my efforts look pretty shit. So it is some kind of GUI editor as you agreed above. But do artists like learning new interfaces? Or new applications? Or are they quite happy and quite comfortable in a world where everything looks like photoshop?
So although the texture itself is not a bitmap, approximations of it that an artist can see and manipulate will be bitmaps. The tools to manipulte it will seems to be a mixture of vector and bitmap(y) operations. Artists don't want to randomly adjust sliders that control the coefficients of an equation until they get what they want. They want to paint some colour operations over a bitmap and have the editor find the some coefficients that do something like this at this particular level of approximation. And yes, I know just how hard that problem is, but it does not mean that a photoshop plugin is impossible. And it would certainly aid adoption of the technology, so a shit version may spring up for other reasons.
Because Peter Jenner is not a musician. But he is someone who makes his money in the music business. This is one of the middlemen standing up and saying, "Hang on a minute, we're all fucked". He doesn't work for a record company, his interest is promoting the band that he is working for so he has a certain amount of leeway in what he says. A record exec is bound by the bottom-line, and owning the distribution channel, and above all *fearing change*. His entire livelyhood is based on preventing change.
Peter Jenner is relevent because he is one of the cogs in the machine that can speak out. His first interest is not preserving the status quo (terrible pun if he did manage them, I hope not...) but in promoting the interest of whichever band he is managing. And if you read what he said in the interview it's startling to hear an industry insider say what he does. "The record companies are fucked, they've raped their own business model to extinction". I've paraphrased him a little, but pg 4 has some interesting stuff.
In particular, his own view of the future that he is selling is not just blanket licensing rights for artists. He is also looking at the supply side - the supply of capital to start bands and promote them to profit. He doesn't use the words, but what he is describing is a stockmarket. Where investors can speculate in talent by buying a share of future profits. This is a really interesting idea for the music industry. A market where return comes directly from popularity.
Indeed. Everyone who might contribute has to weigh the cost of their contribution against the possibility that the innovation will happen anyway without their help. If I need to pay the chip designer $1000 to produce a technology that'll eventually be worth $1 million to me, then as long as there's a greater than 1:1000 chance that the technology will be produced anyway, I should wait for it to happen. However, the more fab owners who also decide to wait, the lower the chances are that the chip will be designed, which flips the odds around and makes it profitable for me to pay the chip designer myself.
...
Sure it would. Just because someone else might also be able to make $1 million but without the initial investment, gaining an "advantage" over me, that doesn't mean I'd turn down the opportunity to make $999,000 in profit by paying the designer myself. I don't need to restrict the use of the invention for that to be in my interest.
OK, if you're talking about an industry where the ratio of investment to sales is 1000:1, and the number of sales is relatively fixed (independent of investment costs) then I can agree with what you're saying. But I think that most industries don't fall into that category. In particular if I invest $X that my competitors don't, then my prices have to be $X/N higher, where N is the number of units that I'm expecting to sell. Where the ratio is high, and therefore X/N is marginal it makes sense to pay a designer and sink the investment cost because it will be made back. My sales won't be adversely affected by the marginal cost added to unit price.
But as the ratio of investment cost to sales decreases, the addition to each unit cost increases. And now the advantage that I get by sitting back and letting someone else do the design work will directly give me his sales. I'll be able to produce the finished item at a cost low enough, compared to the originator, that I can expect to clean-up and take all of the profit.
Competition (in the market) is naturally a winner-takes-all game. Generally these games don't mix well with the alturism being suggested.
First off, you've got it backwards. I wouldn't develop a new chip and then try to find a buyer, I'd shop my chip-designing talent around and find someone who'd pay me to design a new chip. And who'd be willing to pay for it? Anyone who stands to benefit from advancing the state of computers: computer manufacturers, chip fabricators, and consumers.
Lets look at this as a simple piece of Game Theory. Let say there are some fab owners who all make money from chips. New chips would increase consumer demand and so would be could for their business, but who foots the cost? If all the owners club together to pay some designers for a new chip design then nobody wins. Everyone pays the bill, everyone gets the increase in sales. What if everyone but one owner decided to club together to pay the designers? Now one company has an advantage because they don't pay the development costs but can make the chips anyway. The Nash Equilibria are waiting for other people to innovate, and then selling what they come up with.
So in your example, who would want to pay the talent? It wouldn't be in their interest unless they could restrict the use of the invention. Which is where the lack of patents starts to sound like guilds and patronage. Don't get me wrong. Personally I think patents / copyrights / IP in general is a broken solution, but it's are the least broken solution that I've heard so far.
While it's an interesting idea - if ideas don't have any economic value, then why would the ability to produce them? Who cares if you have Einstein himself, even if you don't you still get free access to anything he produces. The only way that the ability would be prized is if there is some restriction on the supply of the ideas. This sounds like the old system of patronage, although in that case the ideas / craft / art was also restricted, giving the patron some incentive to fund their development.
OK. I'm not the OP that you're arguing with but you cme across as someone who thinks they understand Hutter's proof so I've got a few questions for you. From the link that you gave, I've read the background before and skmmed through the ECML paper to remember what he'd done.
The general AI algorithm seems to suffer from two main problem: 1. Specification. The measurement critera is "wooly". If we assume that certain input symbols are nominated rewards from the environment then we are limiting ourselfs to certain types of problen. Not all intellegence problems can be specificed using such a simple class of interaction. In particular what evidence is there that the method generalises to more complex forms of interaction such as deferred and/or partial rewards?
2. Computation. Both the set of p being computed, and the bound t must be increased n order for the search to be universal. Some computable problems may require extreme values of t and/or p, and there is no general strategy to find them.
In short, the general AI model / algorithm doesn't appear to claim anything other than if a solution is computable, and we work in a time-bounded model to avoid non-termination then we will find it sooner or later. This does absolutely nothing to solve the fundamental problems in the field, which are of the nature how. Rather than if.
Well, the box the problem is not the emerge nvidia-drivers. It's the removal of X6.9 and the installation of modular x7.0. There is no way to get nvidia-drivers without an upgrade of the X server. As they're independent packages this is pretty shit, and that is how I got screwed. I've hit a driver bug in the glx that we're using, and needed to check the latest glx to see if Nvidia had fixed it. The fact that I can't check the new driver without reinstalling X suggests that the package system is not working the way that it should. And the fact that the nvidia-glx package has been deleted from tree and replaced by something with a direct dependency on X is unforgivable.
OK so you do seem overly cranky. Some sort of raw nerve there or something?
In a discussion about Ubuntu I think it's perfectly reasonable to say that the recent Gentoo upgrades have been a pain in the ass. For people on the ~amd64 profile it was practically unsupported. Maybe you were luck in your mix of packages that it just worked for you, but it was not a simple case of following written instructions for a lot of people in that position. Lots of ~amd64 packages broke during the upgrade and there were a lot of people who got screwed trying to fix them. Maybe it was their "simple written" instructions that you ended up following? You don't think these guides spring up out of thin air do you? They are generally written by the people who experienced the pain of doing the upgrade first.
And yes, my upgrade was relatively painfree, but I think that is because the x86 and ~x86 profiles are a lot more mainstream than their amd64 counterparts, and because I waited for a couple of weeks at which point there were lots of simple howtos available.
Yeah this bit me this week too. The nvidia-glx package has completely disappeared from the new tree, so after doing an emerge --sync I was screwed. No way to upgrade the driver without doing the whole X thing. No way to rollback to my working tree. It really does bite.
Luckily this box is an amd64 running the x86 profile and so the X upgrade was relatively straightforward. But what ever happened to the idea that Gentoo offered choice?
Not nonsense really. It makes sense to me, which is why I still use Gentoo. There is something reassuring abount a set of command-line tools and forums. Too often a system is borked up too badly to get into the graphical tool. Hmm, actually that might just be my system...
Oddly I know of quite a few people who are planning on dumping Gentoo and switching to Ubuntu. The main reason is the pain of switching "profiles", which is not really supported in Gentoo and can be considered the same as a dist upgrade. The recent modular X headache is another reason, especially when it forces a profile switch to avoid a broken system.
Yeah but come on - this is in Yeovil. Have you ever been in Yeovil on a Friday / Saturday night? It's an absolute arsehole of a town. I would say that fingerprinting is too easy, they just castrate anyone involved in an incident and let natural selection take care of it.
Slashdot Mirror
Number 1 seems a bit wrong. The Via Padlock engine does two things - secure PRNG and it has a Montgomery multiplier. The secure PRNG doesn't help as that's not what the spy process is measuring. The multiplier is called either once, or twice, in the exponentation loop depending on a bit in the key. The leakage channel is the branch that decides whether or not to call the second multiplication. Is the Via Padlock that quick that it prevents the branch choice from leaking though the branch table?
I went to VON (Voice On the Net) 2000 where a lot of people were already exhibiting working systems that implement what the patent describes. The patent is a dead duck. Tons of prior art, and an entire tradeshow devoted to the fact that it was obvious to 'one proficient in the field'. SIP was already around then - designed to interface between different implementations of software phones.
You have to remember that lots of shit patents are filed, but then never used. This will sit in a cupboard forever as it is not worth the money to try and defend it.
No the argument is correct. Although there are lots of sequential dependencies in the game, there is also lots of code that can be parallelised. The parallelisation is limited by the length of the longest sequential chain - this is another phrasing of Ahmdahl's law that the OP quoted. Everything within a timeslice (say a frame) will have a dependency on the IO, and the final output of the frame and sound will be dependent on everything in the timeslice, but this is not a large chain. One frame is typically ouput while the next is being computed (pipelining) - so the two tasks can be performed in parallel if you accept a one-frame latency. This is what most games do when they double-buffer the graphics. So then you are only concerned with the dependencies *within* a timeslice. Is the physics code, the graphics code, the AI code or the sound code dependent on each other within a timeslice? No. Read the article on Steam / Halflife Episode 2 sometime, it explains this coarse-grain parallelism well.
Now you've slipped back to fine-grain parallelism. Here's a clue: do I need to check the bullet against *every* object in the game? Wouldn't that be a bit silly? How about nearby ones only? Oooh, this would need some kind of spatial data-structure like an oct-tree. Then lets split the timestep into two-phases, on the first phase we'll compute all interactions between objects that are well within their spatial sub-division. We'll then broadcast these changes, and then process the areas around intersections. Gosh, a sequetial dependency chain of two steps - that's a constant by the way. And potential parallelism for as many processors as we have sub-divisions (or buckets if you will) in the spatial structure. This is why physics code can be parallelised well.
The comment about electrostatics applies because if you are considering non-local fields of interactions then this approach is borked and you need to consider n-squared interactions. But even then you can simulate in discrete timesteps and trade latency for potential parallelism.
In summary - there are no inter-module sequential dependencies because of double buffering (this is called ping-ponging in the GPGPU community). The intra-module sequential dependencies can be minimised by grouping spatial regions into a single calculation, although as Valve have pointed out this is difficult. And in fact there is 30 years of research into how to solve this problem. For a working example have a look at GPGPU where there is a paper showing a million particle simulation working in parallel on a GPU.
Yeah that's one way to do it. If you can assume the filesystem hasn't been compromised then you could store the executable on disk, otherwise build it at runtime.
RSA isn't the problem. The implementation of RSA on a modern processor is the problem. Moving to another algorithm wouldn't guarantee a lack of side channels. One way around this would be to specialise the algorithm with your own private key. This would unroll all of the loops, and decide the branches statically. If you assume that the machine is not compromised, then this executable could be stored as read-only for your account. If the machine is compromised enough for a non-priviledged process to read your private data then you don't need SBPA - you're toast.
Really? So if I pick the sound code and the graphics code as an example you're telling me that there will be a sequential dependency between them? Just because two things are related does not mean that they are dependent.
Ah, using capitals to reinforce your point. Let me correct your error:
ALL THE OBJECTS THAT DON'T directly INTERACT WITH EACH OTHER.
I'm sure you can tell which word I've added. Now think about this question - do you think that physics simulations on graphics hardware, or the PhysX board are sequential in nature? Although direct interaction implies communication, and hence a sequential dependency in the code, we live in a (mostly) Euclidean universe, and that limits the number of direct interactions. If you're simulating force interactions like electostatics between particles then everything has a direct interaction. But in a game the simulation is of collisions, and gravity. In this case not all objects directly affect each other. Hence they can be clustered into separate groups which can be processed in parallel. Going back to the cell as the example, I'd expect there are more than 8 such groups in every game simulation.
You are assuming that the parallel code and the sequential code have some sort of dependency that force them to run in sequence, otherwise the speedup is 4x for the Cell, rather than 2.9x. Obviously as the 25% bottleneck is reduced, the speedup is greater. The majority of code in a game is parallelisable - although it is not always easy to write it that way. The 25% is more like 2.5% for sequencing IO. All of the graphics code can be parallelised (assuming multiple viewport you can just do software SLI). AI is heavily parallelisable (especially the crappy rule-based search that current games use). Physics simulation is a classic parallel processing problem.
I would expect to see near complete utilisation of the Cell, right around the time that Sony introduces the PS4...
Has to be "wheresmyfusionpoweredflyingcar"
Today. I was really suprised to see a leaflet in the supermarket today for a 100% green energy supplier who will price-match any standard tariff. This means that they're not the cheapest on the market, but they are as cheap as most people pay. This is a uk company, so I don't know if there is something similar for the majority of the us slashdot readership.
For Brits who are interested in 100% green energy at market prices: Ecotricity
Here's a clue. It's not a source-code editor. I've done work in both procedural texuring, and procedural geometry and I use a text editor. But I'm a programmer and my efforts look pretty shit. So it is some kind of GUI editor as you agreed above. But do artists like learning new interfaces? Or new applications? Or are they quite happy and quite comfortable in a world where everything looks like photoshop?
So although the texture itself is not a bitmap, approximations of it that an artist can see and manipulate will be bitmaps. The tools to manipulte it will seems to be a mixture of vector and bitmap(y) operations. Artists don't want to randomly adjust sliders that control the coefficients of an equation until they get what they want. They want to paint some colour operations over a bitmap and have the editor find the some coefficients that do something like this at this particular level of approximation. And yes, I know just how hard that problem is, but it does not mean that a photoshop plugin is impossible. And it would certainly aid adoption of the technology, so a shit version may spring up for other reasons.
Because Peter Jenner is not a musician. But he is someone who makes his money in the music business. This is one of the middlemen standing up and saying, "Hang on a minute, we're all fucked". He doesn't work for a record company, his interest is promoting the band that he is working for so he has a certain amount of leeway in what he says. A record exec is bound by the bottom-line, and owning the distribution channel, and above all *fearing change*. His entire livelyhood is based on preventing change.
Peter Jenner is relevent because he is one of the cogs in the machine that can speak out. His first interest is not preserving the status quo (terrible pun if he did manage them, I hope not...) but in promoting the interest of whichever band he is managing. And if you read what he said in the interview it's startling to hear an industry insider say what he does. "The record companies are fucked, they've raped their own business model to extinction". I've paraphrased him a little, but pg 4 has some interesting stuff.
In particular, his own view of the future that he is selling is not just blanket licensing rights for artists. He is also looking at the supply side - the supply of capital to start bands and promote them to profit. He doesn't use the words, but what he is describing is a stockmarket. Where investors can speculate in talent by buying a share of future profits. This is a really interesting idea for the music industry. A market where return comes directly from popularity.
OK, if you're talking about an industry where the ratio of investment to sales is 1000:1, and the number of sales is relatively fixed (independent of investment costs) then I can agree with what you're saying. But I think that most industries don't fall into that category. In particular if I invest $X that my competitors don't, then my prices have to be $X/N higher, where N is the number of units that I'm expecting to sell. Where the ratio is high, and therefore X/N is marginal it makes sense to pay a designer and sink the investment cost because it will be made back. My sales won't be adversely affected by the marginal cost added to unit price.
But as the ratio of investment cost to sales decreases, the addition to each unit cost increases. And now the advantage that I get by sitting back and letting someone else do the design work will directly give me his sales. I'll be able to produce the finished item at a cost low enough, compared to the originator, that I can expect to clean-up and take all of the profit.
Competition (in the market) is naturally a winner-takes-all game. Generally these games don't mix well with the alturism being suggested.
It doesn't matter how many cameras there are. We all feel safe because we know that Chairman Blair would never abuse the power.
I think you missed the GP's point. A birth certificate is proof that *somebody* was born in the U.S. Proving that *you* are that person is harder...
So in your example, who would want to pay the talent? It wouldn't be in their interest unless they could restrict the use of the invention. Which is where the lack of patents starts to sound like guilds and patronage. Don't get me wrong. Personally I think patents / copyrights / IP in general is a broken solution, but it's are the least broken solution that I've heard so far.
While it's an interesting idea - if ideas don't have any economic value, then why would the ability to produce them? Who cares if you have Einstein himself, even if you don't you still get free access to anything he produces. The only way that the ability would be prized is if there is some restriction on the supply of the ideas. This sounds like the old system of patronage, although in that case the ideas / craft / art was also restricted, giving the patron some incentive to fund their development.
OK. I'm not the OP that you're arguing with but you cme across as someone who thinks they understand Hutter's proof so I've got a few questions for you. From the link that you gave, I've read the background before and skmmed through the ECML paper to remember what he'd done.
The general AI algorithm seems to suffer from two main problem:
1. Specification. The measurement critera is "wooly". If we assume that certain input symbols are nominated rewards from the environment then we are limiting ourselfs to certain types of problen. Not all intellegence problems can be specificed using such a simple class of interaction. In particular what evidence is there that the method generalises to more complex forms of interaction such as deferred and/or partial rewards?
2. Computation. Both the set of p being computed, and the bound t must be increased n order for the search to be universal. Some computable problems may require extreme values of t and/or p, and there is no general strategy to find them.
In short, the general AI model / algorithm doesn't appear to claim anything other than if a solution is computable, and we work in a time-bounded model to avoid non-termination then we will find it sooner or later. This does absolutely nothing to solve the fundamental problems in the field, which are of the nature how. Rather than if.
Well, the box the problem is not the emerge nvidia-drivers. It's the removal of X6.9 and the installation of modular x7.0. There is no way to get nvidia-drivers without an upgrade of the X server. As they're independent packages this is pretty shit, and that is how I got screwed. I've hit a driver bug in the glx that we're using, and needed to check the latest glx to see if Nvidia had fixed it. The fact that I can't check the new driver without reinstalling X suggests that the package system is not working the way that it should. And the fact that the nvidia-glx package has been deleted from tree and replaced by something with a direct dependency on X is unforgivable.
OK so you do seem overly cranky. Some sort of raw nerve there or something?
In a discussion about Ubuntu I think it's perfectly reasonable to say that the recent Gentoo upgrades have been a pain in the ass. For people on the ~amd64 profile it was practically unsupported. Maybe you were luck in your mix of packages that it just worked for you, but it was not a simple case of following written instructions for a lot of people in that position. Lots of ~amd64 packages broke during the upgrade and there were a lot of people who got screwed trying to fix them. Maybe it was their "simple written" instructions that you ended up following? You don't think these guides spring up out of thin air do you? They are generally written by the people who experienced the pain of doing the upgrade first.
And yes, my upgrade was relatively painfree, but I think that is because the x86 and ~x86 profiles are a lot more mainstream than their amd64 counterparts, and because I waited for a couple of weeks at which point there were lots of simple howtos available.
Yeah this bit me this week too. The nvidia-glx package has completely disappeared from the new tree, so after doing an emerge --sync I was screwed. No way to upgrade the driver without doing the whole X thing. No way to rollback to my working tree. It really does bite.
Luckily this box is an amd64 running the x86 profile and so the X upgrade was relatively straightforward. But what ever happened to the idea that Gentoo offered choice?
Not nonsense really. It makes sense to me, which is why I still use Gentoo. There is something reassuring abount a set of command-line tools and forums. Too often a system is borked up too badly to get into the graphical tool. Hmm, actually that might just be my system...
Oh yeah? Tried to "upgrade" from x86 to amd64?
The modular X headache wasn't too bad on x86 and only took a couple of hours. It is practically unsupported on ~amd64...
Oddly I know of quite a few people who are planning on dumping Gentoo and switching to Ubuntu. The main reason is the pain of switching "profiles", which is not really supported in Gentoo and can be considered the same as a dist upgrade. The recent modular X headache is another reason, especially when it forces a profile switch to avoid a broken system.
It works fine in cedgea. It's only $5 difference...
Yeah but come on - this is in Yeovil. Have you ever been in Yeovil on a Friday / Saturday night? It's an absolute arsehole of a town. I would say that fingerprinting is too easy, they just castrate anyone involved in an incident and let natural selection take care of it.