Given that's particularly unhelpful to most people, let me start by pointing out that these beast vary in size.
The molecular size aspect relates to the width, not length of the device, and that width is not dissimilar from a single strand of a polymer (such as polystyrene), but it varies a lot. They mean around the 10's of Angstroms sort of size.
As a side point, atom size is illdefined, and the only reasonable measure is the covalent radii, which are defined in terms of bond length. --
That was a particilurly uninformative article, so heres a little more:
The system involves cabon nanotubes (hence why it got into Science [0]). These devices, sometimes called buckytubes, are semiconductors. Production of a Y junction, which was seen as a prerequisite to transistor like behaviour, was achived about 6 months to a year ago, in various groups (exact time depends on pwhich group and precise nature of the tube).
It's nice to see the somone's made a transistor from this stuff, but it won't replace silicon in the near future.
These are independant devices, and do not lend themselves to lithographic techniques. At least, not by any method thus far discovered.
To make a circuit from these would be like assembuiling it from discrete transistors, but at a near molecular level. --
Not a silly as it might sound, but ed is everywhere.
Vi is not much more than ed on vt100, and if you've ever been confronted with a system that won't accept your using a vt100 capalable terminal, until you edit a config file, that's when ed is useful.
I was first introduced to the joys of ed when I went MUDding. I rapidly shifted to writing them, and ed is the standard built in editor in the LPC MUDs, notably MUDOs.
Course, you could code vi in there, but that don't play nicely with speciist clients.
[As an aside, if you want an intoduction to OOP, grab a copy of the Discworld MUDlib, and fire it up. There is nothing like the sword is a WEAPON is an ITEM is an OBJECT to learn inheriance in moments, and multiple inheritance arrives the instant you want an armoured gauntlet that's also a weapon. After 6 months coding, I was giving OOP tutorials to people studing comp sci, year above me. Which I've never done]
He's not measuring turbidity (at least, not intentionally).
The principle of tghe measurment is more like that of UV-Cis spectroscopy, measuring some set of exitation bands. This is, of course, overlaid onto the turdibity measurment that is also made, but not used.
I don't know enough of the history of computing to argue that on a poin by point basis, so let me explain my reasoning.
Computing became accessable to specialist in the 1950's, possible 1960, depending how you count it.
Quantum mechanics became accessable to the specialist in the 1930's.
Yes, it's a skewed point of view, due to the much easier dissenination of ideas versus hardware.
Of course, another point of view would mention that QM calculations, in reality, require the use of a computer, and thus the field of actually doing things with it must therefore be younger than computing. --
Um, possibly, for a sufficently broad definition of quantum applications. However, by the definition the symposia implicitly imposes, no.
The reason is that those devices do not operate at the level of individual fermions (or bosons). Transistors are bulk devices, needing many electrons. A spintronic-transistor, where the spin of each electron is taken into account (in an analagous fashion to polarised light, might qualifiy, because of the utilisation of the spin of the electrons.
Although, to be honsest, the real reason is that the symposia is for new stuff, and those systems are well understood, without the need to invoke QM constantly.
SQUID's are only used at cryogenic temperatures because they rely on the Josaphen junction, which can only exist in a superconductor. With the advent of high temperature superconductors, the peak is now around 130K (Of the top of my head, not my field). Materials like Magnesium diboride, which superconducts, but probably in a very different fashion to the other ceramics open up a new avenues in that area.
Oh, and I have used a SQUID (a couple of years back), in liquid nitrogen, not liquid helium. --
There is only one linux machine in my group, an old 486 that does print sharing protocol conversion (Samba to lpr).
We have Windows (98 / 2k depending) on the desktop, AIX and Solaris on our number crunchers, HP-UX on the cluster, and an Alpha on Dec Unix in reserve (Old alpha, not that fast). There's an SGI Indy doing print serving and visulisation software in the corner. Oh, and UNICOS is what we use on the CSAR.
We do scienfic computing (Solid state physics), and reliabiltiy is important - particularly when your talking runtimes over around a month. PC hardware doesn't cut it, until you get into the range where UNIX boxes are similar prices. Hence the range of Unixs, and very little linux. --
Not Quantum mechanincs. It's definitly not the first quantum mechanics symposia, given that happen in the 1930's, with less formal meetings before hand.
Oh, and given I was at a quantum mechanics comference last month, I realy hope this isn't the first.
To explain the nomenclature:
Quantum mechanics is the mathematical framework used for dealing with fermions - i.e. the Schodinger equation and all it's baggage. This includes all the calculateion schemes, such as Hartree-Fock and the Density function theories that allow you to actually solve the Schodinger equation, (Or, rather, give approximate solutions). These are useful, old, and not cool.
Quantum applications, is ment to reffer to any system where the quantum properties of a system are exploited for some macroscopic function. The ubercool quantum computing is one example. These are not yet usable, very new, and considerably cooler than the quantum mechanics required to describe them.
Maybe I'm being a little pedantic here, but if your really interested in this, look at the correct areas would be a useful move.
Remeber that Quantum mechanics is pretty much twice the age of computing, give or take. --
If they were going for straightforawrd comparison, your correct.
However, consider this: You can recognise a song, from, say, 10-20 seconds of it?
How?
If the way that _you_ recognise the song (which is indepentant of encoding, bit rate etc), is implemented in a computer, then provided that a human can ID the song, no matter what it is, so can a computer.
The way to do this would be to use some extremely harsh psycho acoustics - like those used in MP3 or Vorbis, but throw away a lot. Given Naptsar is MP3 only, I'd extract the key parts of the song that exist at really low bitrates, and assume that they must be present at all bit rates. Then look for those.
It's difficult, yes. Artifacts are a big issue too, but I beleive it's perfectly possible.
Note that this approach does not inherently use a neural net, but instead attempt to mimic the pathway between the ears and the cereral cortex. This reduces the problem to one asking if two files, of known representation, are similar, a difficult, but known to be soluble, problem. --
If they presented the information, whiout a liscence, they have the rights given to them by basic copyright law (In brief, you can use it for personal use only)
A liscence allows the copyright holder to modify those rights.
The GPL does _not_ protect anything - it gives the liscencee _more_ rights.
And, further, if I recall correctly, he original data _was_ GPL'd. --
When silicon is deposited on top of a substrate with atoms spaced farther apart, the atoms in silicon stretch to line up with the atoms beneath, stretching -- or "straining" -- the silicon.
So, you need to build the silicon on top of a substrate, with a similar crystal surface, but a larger lattice parameter(s). Then grow the silicon on top of it by some technique that maintains atomic level consitancy between the layers.
This is difficult to do - your basically talking something along the lines of silicon deposited by some for of epitaxial growth - and for thick layers that's a timeconsuming process. And thus expensive.
One thing that was not mentioned was the cost of this trick - how does it compare with germanium or gallium arsnide? (Ok, projected to compare?).
I think that, baring some lucky find, this is going to be more expensive than the befefits, for general use.
Not only that, but the interdiffusion coefficents of a strained material are, in general, faster than for the an unstrained material, so this will decrease the lifetime of the devices.
Whilst I don't know exactly what system this uses, my PhD is in magnetic intereactions in systems of reduced dimensionality, so this will work, and wont be far off.
The problem with increasing data density is that neighbouring bits get closer together. There is an interaction between the bits, and the closer they get, the stronger it is.
The antiferromagnetic coupling introduced by the ruthenium (as metalic layer, btw, not as dust), make it more difficult for a bit to be flipped by a neighbour. This allows neigbouring bits to be place closer together, because they can now withstand higher field while maintaining data integrety. --
I urge you to go and look at the LKML, or, more likely, KT. Specifically, the thread about X15 compared to TUX (the first is user space, the second is the kernel web _accelerator_ you mentioned).
X15 is faster. Not much, but slightly.
It's an issue of design, thorwing it into the kernel does not, inherently, make it faster, in general. --
I'm pretty sure you can't copyright factual infomation
That is corect, (UK wise at any rate) however...
The act of compling the data into the database is protected. So, whilst Gracenote cannot stop anyone else from makeing a database with the same information in it, the can stop someone taking Gracenotes database and selling it on.
Legally, a database is more than just a collection of data. Which is fair enough, if you've spent time and money compling the DB, you don't want all that effort pinched. However, if someone else puts the effort in, you can't stop them profiting of the results.
For legal advice, see a lawyer in your own juristriction.
I know of no commercial motherboards, or chip sets for that matter, that have more than one AGP slot. For two good reasons..
1: Why would you want to have more than one graphics card?
Bearing in mind AGP is designed to optimise 3D texture transfer, the argument of a 2d and 3d card is irrelevnt, and pointless with modern cards.
If the system was to switch between work and games, then, and only then, can I see an argument for two cards. But for a games specific machine, nope.
This means that active communication can only occur between two AGP agents that reside on the interface... Attaching more then two devices to the interface is not precluded, as long as there is only one active master and one active target
In other words, one AGP card (the target is the motherboard, master is the card) has to be switched off during AGP transfers.
...it's just that knowing you look good is the biggest self-confidence booster.
I have to respectfully disagree. Being comfortable with your appearence is what counts. If being comfortable means blending in, then fine, go and do that.
But, speaking as someone who is disfigured, it doesn't matter what you wear. Just that you are comforatable with it. I wear black jeans, and (plain) black t-shirt, because that's what I'm confortable in. Wearing trendy clothes wouldn't stop people from pointing and laughing at me, so I don't even start.
You don't have to look good to project confidence. I found that taking up Aikido did a lot more for me projecting confidence than anything else.
Really, the most important thing to add is don't let someone drag you out to get new clothes. That doesn't work. It only works if you go yourself. You don't have to.
1. What is lzip?
Lzip is the most advanced file compression utility ever conceived. It is literally years ahead of gzip (though admittedly gzip was around first), and makes use of mathematical transforms the bzip developers have never even heard of.
So the lzip people know, exactly, the contents of the bzip developers heads? They'll be telepaths then.
2. What makes lzip different from gzip/bzip2?
The second is that the performance is vastly improved.
Tell me it takes longer to MP3 encode than it does to gzip or bzip2. (MP3 being the most used lossy codec, to slashdotters).
We're talking about a constant-time algorithm that can reduce a file down to 0% of its original size. What's not to like?
Contsant time - not linear. In otherwords, give it a large enough file, and it can compress it before it reads it all off the disk.
Compress down to 0% - I reffer the reader to the comp.compression FAQ, point 73. It's impossible.
3. What do you mean I can't restore my files?
On the reassuring side, it is important to note that the compression algorithm used by lzip only discards the unimportant data.
An algotithm that works on generic files, but can identify the important data in them. Guess all the compression research just got made useles then - identfying what parts of the data is important is impossible, only a human can do that [0].
8. What is the Lessiss-Moore algorithm?
It utilizes a two-pass bit-sieve to first remove all unimportant data from the data set. Lzip implements this quiet effectively by eliminating all of the 0's.
Uh-hu. Removes all the 0's. Do I really need to elaborate on that one.
Oh, and thier liscence is the FOOL liscence. Go figure.
[0] In MP3, a human came up with the psychoacoutstic model. The codec just applies that model. --
The presentation was about features that XFS developers would like to see in the kernel, and why.
It means no more than suggestions from Oracle. If you look at the things they suggest, they also give reasons for them. Remember that XFS is a fairly maure journeling FS, compared to Reiserfs which started for Linux.
Resierfs is in the Kernel, last I checked XFS wasn't yet. Both, and ext3, will be around for a while. --
2) You took steps to promote thier growth, knowing what they were.
If you had a liscence to grow THC free marijuana, but a couple fo plnats were the wild type, they you would have an excuse. However, as it's easy ti ID marijuana, that doesn't hold.
GM rapeseed looks identical to proper rapeseed. The farmer did not, and could not, identify the difference, and treated it like the expected crop.
Slashdot Mirror
Read the article.
.NE. i) (x_j - x_i) / (x_j - x_i)^3
With refference to the calculatiosn they are doing, they are simply doing
G * m_i * SumOverAll(j
They are doing this by custom hardware.
This is not a general purpose computer.
Despite what the blurb said, there are 96 independant units doing the calculation, in each machine, to get the 32 TFlops across the system.
There is a picture of an earlier model, which is about the size of one of my filing cabinets.
Remeber these are scientists, not marketing, making those claims. They expect to be asked to justify them - and they have.
--
Yes - carbon nanotube sized molecule.
Given that's particularly unhelpful to most people, let me start by pointing out that these beast vary in size.
The molecular size aspect relates to the width, not length of the device, and that width is not dissimilar from a single strand of a polymer (such as polystyrene), but it varies a lot. They mean around the 10's of Angstroms sort of size.
As a side point, atom size is illdefined, and the only reasonable measure is the covalent radii, which are defined in terms of bond length.
--
That was a particilurly uninformative article, so heres a little more:
The system involves cabon nanotubes (hence why it got into Science [0]). These devices, sometimes called buckytubes, are semiconductors. Production of a Y junction, which was seen as a prerequisite to transistor like behaviour, was achived about 6 months to a year ago, in various groups (exact time depends on pwhich group and precise nature of the tube).
It's nice to see the somone's made a transistor from this stuff, but it won't replace silicon in the near future.
These are independant devices, and do not lend themselves to lithographic techniques. At least, not by any method thus far discovered.
To make a circuit from these would be like assembuiling it from discrete transistors, but at a near molecular level.
--
Not a silly as it might sound, but ed is everywhere.
Vi is not much more than ed on vt100, and if you've ever been confronted with a system that won't accept your using a vt100 capalable terminal, until you edit a config file, that's when ed is useful.
I was first introduced to the joys of ed when I went MUDding. I rapidly shifted to writing them, and ed is the standard built in editor in the LPC MUDs, notably MUDOs.
Course, you could code vi in there, but that don't play nicely with speciist clients.
[As an aside, if you want an intoduction to OOP, grab a copy of the Discworld MUDlib, and fire it up. There is nothing like the sword is a WEAPON is an ITEM is an OBJECT to learn inheriance in moments, and multiple inheritance arrives the instant you want an armoured gauntlet that's also a weapon. After 6 months coding, I was giving OOP tutorials to people studing comp sci, year above me. Which I've never done]
--
He's not measuring turbidity (at least, not intentionally).
The principle of tghe measurment is more like that of UV-Cis spectroscopy, measuring some set of exitation bands. This is, of course, overlaid onto the turdibity measurment that is also made, but not used.
--
Good arguments.
I don't know enough of the history of computing to argue that on a poin by point basis, so let me explain my reasoning.
Computing became accessable to specialist in the 1950's, possible 1960, depending how you count it.
Quantum mechanics became accessable to the specialist in the 1930's.
Yes, it's a skewed point of view, due to the much easier dissenination of ideas versus hardware.
Of course, another point of view would mention that QM calculations, in reality, require the use of a computer, and thus the field of actually doing things with it must therefore be younger than computing.
--
Um, possibly, for a sufficently broad definition of quantum applications. However, by the definition the symposia implicitly imposes, no.
The reason is that those devices do not operate at the level of individual fermions (or bosons). Transistors are bulk devices, needing many electrons. A spintronic-transistor, where the spin of each electron is taken into account (in an analagous fashion to polarised light, might qualifiy, because of the utilisation of the spin of the electrons.
Although, to be honsest, the real reason is that the symposia is for new stuff, and those systems are well understood, without the need to invoke QM constantly.
SQUID's are only used at cryogenic temperatures because they rely on the Josaphen junction, which can only exist in a superconductor. With the advent of high temperature superconductors, the peak is now around 130K (Of the top of my head, not my field). Materials like Magnesium diboride, which superconducts, but probably in a very different fashion to the other ceramics open up a new avenues in that area.
Oh, and I have used a SQUID (a couple of years back), in liquid nitrogen, not liquid helium.
--
True enough, I forgot about bosons. It's what I get for doing all-electron calculatiosn too much.
I aught to go and look over boson calculations again..
--
There is only one linux machine in my group, an old 486 that does print sharing protocol conversion (Samba to lpr).
We have Windows (98 / 2k depending) on the desktop, AIX and Solaris on our number crunchers, HP-UX on the cluster, and an Alpha on Dec Unix in reserve (Old alpha, not that fast). There's an SGI Indy doing print serving and visulisation software in the corner. Oh, and UNICOS is what we use on the CSAR.
We do scienfic computing (Solid state physics), and reliabiltiy is important - particularly when your talking runtimes over around a month. PC hardware doesn't cut it, until you get into the range where UNIX boxes are similar prices. Hence the range of Unixs, and very little linux.
--
Not Quantum mechanincs. It's definitly not the first quantum mechanics symposia, given that happen in the 1930's, with less formal meetings before hand.
Oh, and given I was at a quantum mechanics comference last month, I realy hope this isn't the first.
To explain the nomenclature:
Quantum mechanics is the mathematical framework used for dealing with fermions - i.e. the Schodinger equation and all it's baggage. This includes all the calculateion schemes, such as Hartree-Fock and the Density function theories that allow you to actually solve the Schodinger equation, (Or, rather, give approximate solutions). These are useful, old, and not cool.
Quantum applications, is ment to reffer to any system where the quantum properties of a system are exploited for some macroscopic function. The ubercool quantum computing is one example. These are not yet usable, very new, and considerably cooler than the quantum mechanics required to describe them.
Maybe I'm being a little pedantic here, but if your really interested in this, look at the correct areas would be a useful move.
Remeber that Quantum mechanics is pretty much twice the age of computing, give or take.
--
If they were going for straightforawrd comparison, your correct.
However, consider this: You can recognise a song, from, say, 10-20 seconds of it?
How?
If the way that _you_ recognise the song (which is indepentant of encoding, bit rate etc), is implemented in a computer, then provided that a human can ID the song, no matter what it is, so can a computer.
The way to do this would be to use some extremely harsh psycho acoustics - like those used in MP3 or Vorbis, but throw away a lot. Given Naptsar is MP3 only, I'd extract the key parts of the song that exist at really low bitrates, and assume that they must be present at all bit rates. Then look for those.
It's difficult, yes. Artifacts are a big issue too, but I beleive it's perfectly possible.
Note that this approach does not inherently use a neural net, but instead attempt to mimic the pathway between the ears and the cereral cortex. This reduces the problem to one asking if two files, of known representation, are similar, a difficult, but known to be soluble, problem.
--
This is deliberate.
With open source liscences, there are two main schools of thought:
GPL: Not going anywhere near commercial code. If you want to use it, either GPL or don't. Harsh, but not as harsh as many commercial liscences.
LGPL/BSD: You can use it in commercial code, either verbatim chunks (BSD), or as a distinct, linked, entity.
It's the programmers choice. The GPL may limit it's audience, but does it limit it's audience more, or less, than a typical commercial liscence?
--
Incorrent, boy wonder.
If they presented the information, whiout a liscence, they have the rights given to them by basic copyright law (In brief, you can use it for personal use only)
A liscence allows the copyright holder to modify those rights.
The GPL does _not_ protect anything - it gives the liscencee _more_ rights.
And, further, if I recall correctly, he original data _was_ GPL'd.
--
So, you need to build the silicon on top of a substrate, with a similar crystal surface, but a larger lattice parameter(s). Then grow the silicon on top of it by some technique that maintains atomic level consitancy between the layers.
This is difficult to do - your basically talking something along the lines of silicon deposited by some for of epitaxial growth - and for thick layers that's a timeconsuming process. And thus expensive.
One thing that was not mentioned was the cost of this trick - how does it compare with germanium or gallium arsnide? (Ok, projected to compare?).
I think that, baring some lucky find, this is going to be more expensive than the befefits, for general use.
Not only that, but the interdiffusion coefficents of a strained material are, in general, faster than for the an unstrained material, so this will decrease the lifetime of the devices.
Interesting idea, though.
--
Whilst I don't know exactly what system this uses, my PhD is in magnetic intereactions in systems of reduced dimensionality, so this will work, and wont be far off.
The problem with increasing data density is that neighbouring bits get closer together. There is an interaction between the bits, and the closer they get, the stronger it is.
The antiferromagnetic coupling introduced by the ruthenium (as metalic layer, btw, not as dust), make it more difficult for a bit to be flipped by a neighbour. This allows neigbouring bits to be place closer together, because they can now withstand higher field while maintaining data integrety.
--
I urge you to go and look at the LKML, or, more likely, KT. Specifically, the thread about X15 compared to TUX (the first is user space, the second is the kernel web _accelerator_ you mentioned).
X15 is faster. Not much, but slightly.
It's an issue of design, thorwing it into the kernel does not, inherently, make it faster, in general.
--
That is corect, (UK wise at any rate) however...
The act of compling the data into the database is protected. So, whilst Gracenote cannot stop anyone else from makeing a database with the same information in it, the can stop someone taking Gracenotes database and selling it on.
Legally, a database is more than just a collection of data. Which is fair enough, if you've spent time and money compling the DB, you don't want all that effort pinched. However, if someone else puts the effort in, you can't stop them profiting of the results.
For legal advice, see a lawyer in your own juristriction.
--
1: Why would you want to have more than one graphics card?
Bearing in mind AGP is designed to optimise 3D texture transfer, the argument of a 2d and 3d card is irrelevnt, and pointless with modern cards.
If the system was to switch between work and games, then, and only then, can I see an argument for two cards. But for a games specific machine, nope.
2: To quote the AGP 2.0 specification (footnote 3, page 27)
In other words, one AGP card (the target is the motherboard, master is the card) has to be switched off during AGP transfers.
There is a reason AGP is a Port, not a Bus.
--
IANAL
It's not quite that bad - If you can prove you wrote the code, it becomes Microsoft's responsability to prove that you sent it over Passport stuff.
It is not up to you to prove you didn't.
Admitingly, if you belive that they would falsify evidence, then that's a moot point. But barring that, the onus of proof is on them, not you.
--
...it's just that knowing you look good is the biggest self-confidence booster.
I have to respectfully disagree. Being comfortable with your appearence is what counts. If being comfortable means blending in, then fine, go and do that.
But, speaking as someone who is disfigured, it doesn't matter what you wear. Just that you are comforatable with it. I wear black jeans, and (plain) black t-shirt, because that's what I'm confortable in. Wearing trendy clothes wouldn't stop people from pointing and laughing at me, so I don't even start.
You don't have to look good to project confidence. I found that taking up Aikido did a lot more for me projecting confidence than anything else.
Really, the most important thing to add is don't let someone drag you out to get new clothes. That doesn't work. It only works if you go yourself. You don't have to.
And yes, I have a girlfriend.
--
In order, from the FAQ:
1. What is lzip?
Lzip is the most advanced file compression utility ever conceived. It is literally years ahead of gzip (though admittedly gzip was around first), and makes use of mathematical transforms the bzip developers have never even heard of.
So the lzip people know, exactly, the contents of the bzip developers heads? They'll be telepaths then.
2. What makes lzip different from gzip/bzip2?
The second is that the performance is vastly improved.
Tell me it takes longer to MP3 encode than it does to gzip or bzip2. (MP3 being the most used lossy codec, to slashdotters).
We're talking about a constant-time algorithm that can reduce a file down to 0% of its original size. What's not to like?
Contsant time - not linear. In otherwords, give it a large enough file, and it can compress it before it reads it all off the disk.
Compress down to 0% - I reffer the reader to the comp.compression FAQ, point 73. It's impossible.
3. What do you mean I can't restore my files?
On the reassuring side, it is important to note that the compression algorithm used by lzip only discards the unimportant data.
An algotithm that works on generic files, but can identify the important data in them. Guess all the compression research just got made useles then - identfying what parts of the data is important is impossible, only a human can do that [0].
8. What is the Lessiss-Moore algorithm?
It utilizes a two-pass bit-sieve to first remove all unimportant data from the data set. Lzip implements this quiet effectively by eliminating all of the 0's.
Uh-hu. Removes all the 0's. Do I really need to elaborate on that one.
Oh, and thier liscence is the FOOL liscence. Go figure.
[0] In MP3, a human came up with the psychoacoutstic model. The codec just applies that model.
--
The presentation was about features that XFS developers would like to see in the kernel, and why.
It means no more than suggestions from Oracle. If you look at the things they suggest, they also give reasons for them. Remember that XFS is a fairly maure journeling FS, compared to Reiserfs which started for Linux.
Resierfs is in the Kernel, last I checked XFS wasn't yet. Both, and ext3, will be around for a while.
--
You say this as if it were something now. If you have a bank account, your money has been nothing more than a number in a database for tens of years.
Actually, economic historians belive that scince the days of the first bank, you money has been nothing more than a number in a database.
Essentially, with the first money lenders, money became detatched from tangible goods. Not fully detatched, but somewhat.
That would put it at over 100 years ago. (Possibly a lot more, I'm being safe with 100 years).
--
Yes, because:
1) You clearly identified what they were.
2) You took steps to promote thier growth, knowing what they were.
If you had a liscence to grow THC free marijuana, but a couple fo plnats were the wild type, they you would have an excuse. However, as it's easy ti ID marijuana, that doesn't hold.
GM rapeseed looks identical to proper rapeseed. The farmer did not, and could not, identify the difference, and treated it like the expected crop.
And that's the difference
--
Right, now I'm worried.
This slight seed mixing is exactly what they found in Italy.
AP story
Reuters story
--