how is it legal and supports FastTrack network at the same time?
It is perfectly legal to use the FastTrack network without authorisation from Sharman... they don't own the computers that constitute the network, they just own the software that is usually used to run it.
If you write your own re-implementation by reverse engineering KaZaA to determine how it works, that is perfectly legal. For the same reason that, for instance, Wine is a perfectly legal piece of software.
So, basically, Morpheus has a re-implementation of FastTrack from scratch.
What is possibly a little concerning is that it appears to be a download-only implementation...
In each corpus, create a new 'virtual' message (which will be calculated during the retraining phase). The contents of this message are 1 occurrence each of each word that occurs in the other corpus.
This will mean that when a word has only occurred a few times in one corpus, yet still never at all in the other, its score will be close to 0.5, rather than the 0.9999 (or similar) that traditional techniques would assign to it. Only as the corpus sizes increase and the number of messages containing the word increase with them will the probability deviate towards either 0 or 1 (neither of which will it ever quite reach).
I believe this is the technique Paul Graham is feeling towards in Better Bayesian Filtering when he says "There are theoretical arguments for giving these two tokens substantially different probabilities".
I hope to hell they're fishing for non-bouncing addresses, because at the moment any email which SpamAssassin says is spam, I bounce.
Don't ever do that, all spam has forged headers. You're just making life hard on someone who had their address sold
Yes, those of us who do this are aware of this problem.
However, we feel that it is more important that people who are trying to send legitimate messages that are caught as false positives by filters are aware that their message did not get through.
As someone who has frequently suffered as the forged sender of large mailshots in the past, I can quite understand how frustrating it is. But then, I tend to get thousands of copies of the virus-du-jour too.
4. A press release with the statement 'we have a computer program that shows Martian time on a PDA' is a lot less captivating than 'we got a watch manufacturer to make us Martian watches - you can get them from XXXX.'
Don't even get me started on old, mechanically controlled automatic transmissions. These collections of flywheels, springs, valves, gaskets, and hydraulic clutches are practically works of art.. The result? A lousy transmission which breaks all the time. My friend and I drove an '85 K-5 Blazer (4x4) to Mexico and went thru two automatic transmissions in one trip, I shit you not. The first replacement didn't work, and ended up partially shredding itself. To its credit, AAMCO replaced it free of charge.
Saying you drove to Mexico is useless. I mean, if you started in Texas, that would be particularly bad. If you started in New York, it would still be bad, but understandable. And, of course, whatever you buy, you occasionally ge3t faulty goods that don't work for very long. The manufacturer replacing it is basically an admission that they screwed up manufacturing it. So, essentially, you took a trip of an unspecified length, during which a transmission of unspecified age broke. You replaced it with one that had been badly manufactured. Bad luck.
Meanwhile, I have seen other automatic transmissions last out 70,000 miles before needing a replacement.
My current car only needs its oil changed every 10,000 miles, for goodness sake.
So do some (although not all) 1980s luxury European cars (e.g. Mercedes, Jaguar, BMW). The rest tend to go 7,000, which isn't a huge difference, if you ask me.
When they have a technology that can do what Google Answers does, then it'll be safe for them to IPO, but not before.
When they have a technology that can do what Google Answers does, then it'll be time to sell all your shares in everything and go and live out in the wilderness for a few years, just in case...
Pulling some electric car numbers out of the air, a $3000 lead-acid battery pack stores 10 kWHr per charge times 300-500 charges or 60 cents to a dollar per kilowatt hour.
Those numbers really do seem "pulled out of the air" to me.
I've recently been investigating electric car battery systems, and the pack I've been looking at would cost about UKP1000 (about $1600 US), stores about 12kWhr, and carries an estimated lifespan of 1500 cycles at 75% depth of discharge. That works out to 1600 / 0.75*12*1500 = approx 12c per kilowatt hour.
This means that even if increased the gas engine efficiency to 0 cents per kWHr (through smaller engine run at peak efficiency to only charge batteries) but ran all the power flows through the lead acid batteries, you costs would increase from 40 cents to 60 cents per kWHr delivered to the wheels.
If you used this kind of arrangement, rather than the deep-cycling normally experienced with electric cars, you would probably get approximately two to three times the lifetime out of the batteries, because you would rarely deep cycle them. You also wouldn't need the capacity to be as high, meaning you can use cheaper batteries.
I'm not certain of the regulatory environment in the USA, but in some countries, diesel used to charge batteries is not taxed, whereas diesel used to directly drive a car is taxed at a fairly high rate. That can make this kind of setup _much_ cheaper to run.
Title 17 - COPYRIGHTS Chapter 2 - Copyright Ownership and Transfer [...] Everything contributed by Caldera etc. to Linux under the UNSIGNED nonexclusive GPL prior to the present SCO's purchase (transfer of ownership) from Novell is eliminated by sec 205(e) from the protection of the GPL under the Copyright Act.
1. The GPL is not a transfer of copyright ownership, which is what this law refers to. It is a licence. Licences do not have to be signed, otherwise you'd need Bill Gates' signature in order to run a copy of Windows.
But it may only last untill EUCD is implemented...
The UK EUCD implementation grants specific requirements that mean you must be allowed access to an unencrypted copy of the data if you have a legitimate reason for needing it... unforuntately the procedure for enforcement of this is ridiculously complicated (you must complain to the home secretary, IIRC).
I'm not sure how it'll work in norway, but I'd hope there would be some similar provision.
However face recognition might not be the perfectly viable solution since it has produced too many false positives in the past.
False positives aren't too much of a problem here. I think face recognition has about a 1/1000 false positive rate, which is a killer for crowd recognition, but would be entirely acceptable for this application (the result would be that 1 time in a thousand somebody tried to use someone else's passport, they wouldn't be caught).
False negatives are what we need to worry about in this situation... I don't know much about the statistics here. Anyone have any idea?
Except for the fact that nobody would accept your document as authoritative, because nobody has any authority to produce such a document, except perhaps the X Consortium, who have no interest in producing one, and would almost certainly be unable to agree with the details if there were.
There is generally speaking consisteny in Windows beause MS is accepted as having authority to declare how user interfaces should work there; the same is true of Apple having authority over OSX. Nobody has that authority over X, so it splinters into individual groups (GNOME, KDE, CDE, etc.) who have their own guidelines. So you then need to learn multiple guidelines in order to be able to use the same apps. And there are apps (e.g. emacs) which don't follow any of them.
When I was working for a major global firm, and we dealt with small closed source development companies, we always had code escrow agreements. If the vendor went out of business or dropped support of the product, we had the ability to get the source and support it ourselves.
Now, I can assure you, you can only normally get this kind of agreement if you have a lot of spare cash to throw around. Throwing around IP rights like that limits a business that is being wound up from selling its products to another business, and that is something that smart companies will not do except for a good price.
I also wouldn't wager anybody's chances of making this happen with any kind of mass-marketed software, even if it is aimed at a specialised market. For instance: my company uses a piece of proprietary software to analyse web pages for search engine keyword placement as a method to attempt to improve rankings. I severely doubt that in any way could we persuade the company that wrote this software to agree to the sort of terms you are suggesting, unless we offered to pay them substantially more money than that software is worth to us.
And at the end of all this, what do you end up with? Well, pretty much the same as if you started using an open source product that stopped being developed.
that make differences in behaviour even more annoying - should I use Alt+F4 or Ctrl+W to close an application window, for instance?
I have never seen an application for which Alt+F4 doesn't work, and if there are any, it would be a violation of the Microsoft UI guidelines (Available here). Ctrl+W doesn't appear to actually be standardised, but as a de-facto standard often closes the document you're working with, rather than the entire application. Hope this helps!
Maybe not. But if it ever came to court, it wouldn't be Linus that had to decide. And, by the straightest possible reading of the GPL, trying to divine the original intention of it when it was written (which is, I think, how a court would approach it), such things are probably allowed.
It all rests on what a derivitive work is.
A case to consider: if any software that depends on any other's internal workings in order to funtion correctly is a derivitive work of the other, then is my Windows software a derivitive of Windows, and if so, when will Microsoft be starting proceedings against me? My software "needs to know intimately" the Windows APIs otherwise it wouldn't work, and it "gets linked to [them] when it's loaded", using pretty much exactly the same mechanisms as a Linux kernel module.
what if someone wrote another standard driver interface, separate from the kernel interface, wrote a device driver which implemented that and then wrote a GPL'd interface wrapper which translated the Linux module interface to that of the new standard?
Obviously, the wrapper interface code is now a derived work. However, does it also mean that because the new driver which uses a code interface which the GPL'd wrapper implements now is tainted by the GPL?
From the GPL:
If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it.
Therefore the GPL does not apply to the original devie driver unless and until it is distributed alongside the GPL'd interface wrapper linked together in a single compiled image (by my reading, although IANAL).
Also, does the driver become a derived work if the person who writes it initially does so to get some hardware working on his Linux box, rather than his other box which runs ObsureOS which also implements this standard device driver interface but the person hasn't installed the hardware on that machine yet?
I don't think that would make any difference: as long as the driver was useful without the GPL'd code it could "reasonably [be] considered [an] independent and separate [work] in [itself]" to borrow the GPL's phrasing and adjust for the singular rather than the plural.
Dynamically linking to a GPL'd library does not _necessarily_ cause a program to become licensed under the GPL.
It has long been held in copyright law (through case history) that using an API does not, of itself, mean that the program that uses it is a derivitive of an implementation of that API.
This would be especially true if two implementations of the same API existed, which is probably the case in the nVidia drivers that were being discussed; there is a small driver module (whose source code is available) that merely provides an API layer between the kernel and another program (no source code available). My guess is that that program is simply a recompilation, with no source code changes, of the equivalent program under Windows, which would use a different compatibility layer.
It is therefore plainly not a derivitive of the Linux compatibility layer, because it could be used separately from it. And by the terms of the GPL, as it is a separable program that doesn't depend entirely upon the other, it does not have to be GPL licensed.
1. The specifications probably change with each new card release 2. One of the parent posts gave an estimate of (IIRC) 1.8 Million lines of code in the library that they aren't giving out the source to. Are you volunteering to duplicate that?
Slashdot Mirror
how is it legal and supports FastTrack network at the same time?
It is perfectly legal to use the FastTrack network without authorisation from Sharman... they don't own the computers that constitute the network, they just own the software that is usually used to run it.
If you write your own re-implementation by reverse engineering KaZaA to determine how it works, that is perfectly legal. For the same reason that, for instance, Wine is a perfectly legal piece of software.
So, basically, Morpheus has a re-implementation of FastTrack from scratch.
What is possibly a little concerning is that it appears to be a download-only implementation...
A solution to this problem:
In each corpus, create a new 'virtual' message (which will be calculated during the retraining phase). The contents of this message are 1 occurrence each of each word that occurs in the other corpus.
This will mean that when a word has only occurred a few times in one corpus, yet still never at all in the other, its score will be close to 0.5, rather than the 0.9999 (or similar) that traditional techniques would assign to it. Only as the corpus sizes increase and the number of messages containing the word increase with them will the probability deviate towards either 0 or 1 (neither of which will it ever quite reach).
I believe this is the technique Paul Graham is feeling towards in Better Bayesian Filtering when he says "There are theoretical arguments for giving these two tokens substantially different probabilities".
I hope to hell they're fishing for non-bouncing addresses, because at the moment any email which SpamAssassin says is spam, I bounce.
Don't ever do that, all spam has forged headers. You're just making life hard on someone who had their address sold
Yes, those of us who do this are aware of this problem.
However, we feel that it is more important that people who are trying to send legitimate messages that are caught as false positives by filters are aware that their message did not get through.
As someone who has frequently suffered as the forged sender of large mailshots in the past, I can quite understand how frustrating it is. But then, I tend to get thousands of copies of the virus-du-jour too.
Home has 'fast user switching', which is, IMHO, just about the only XP feature that is worth having over Win2K.
You forgot:
4. A press release with the statement 'we have a computer program that shows Martian time on a PDA' is a lot less captivating than 'we got a watch manufacturer to make us Martian watches - you can get them from XXXX.'
Saying you drove to Mexico is useless. I mean, if you started in Texas, that would be particularly bad. If you started in New York, it would still be bad, but understandable. And, of course, whatever you buy, you occasionally ge3t faulty goods that don't work for very long. The manufacturer replacing it is basically an admission that they screwed up manufacturing it. So, essentially, you took a trip of an unspecified length, during which a transmission of unspecified age broke. You replaced it with one that had been badly manufactured. Bad luck.
Meanwhile, I have seen other automatic transmissions last out 70,000 miles before needing a replacement.
My current car only needs its oil changed every 10,000 miles, for goodness sake.
So do some (although not all) 1980s luxury European cars (e.g. Mercedes, Jaguar, BMW). The rest tend to go 7,000, which isn't a huge difference, if you ask me.
About twelve seconds after Photoshop Preemptive First Strike.
That's a great picture! Thanks :-)
When they have a technology that can do what Google Answers does, then it'll be safe for them to IPO, but not before.
When they have a technology that can do what Google Answers does, then it'll be time to sell all your shares in everything and go and live out in the wilderness for a few years, just in case...
Much scarier than Y2K.
Are they using http?
Well, I don't know much about other P2P applications, but I know gnutella uses http for its file transfers.
Pulling some electric car numbers out of the air, a $3000 lead-acid battery pack stores 10 kWHr per charge times 300-500 charges or 60 cents to a dollar per kilowatt hour.
Those numbers really do seem "pulled out of the air" to me.
I've recently been investigating electric car battery systems, and the pack I've been looking at would cost about UKP1000 (about $1600 US), stores about 12kWhr, and carries an estimated lifespan of 1500 cycles at 75% depth of discharge. That works out to 1600 / 0.75*12*1500 = approx 12c per kilowatt hour.
This means that even if increased the gas engine efficiency to 0 cents per kWHr (through smaller engine run at peak efficiency to only charge batteries) but ran all the power flows through the lead acid batteries, you costs would increase from 40 cents to 60 cents per kWHr delivered to the wheels.
If you used this kind of arrangement, rather than the deep-cycling normally experienced with electric cars, you would probably get approximately two to three times the lifetime out of the batteries, because you would rarely deep cycle them. You also wouldn't need the capacity to be as high, meaning you can use cheaper batteries.
I'm not certain of the regulatory environment in the USA, but in some countries, diesel used to charge batteries is not taxed, whereas diesel used to directly drive a car is taxed at a fairly high rate. That can make this kind of setup _much_ cheaper to run.
The letters include an olive branch as well as a threat.
;-)
Anyone else think "Pay us, or we'll hit you with olive branches!"? Could be quite painful - I'd comply if I received a letter like that...
Title 17 - COPYRIGHTS
Chapter 2 - Copyright Ownership and Transfer
[...]
Everything contributed by Caldera etc. to Linux
under the UNSIGNED nonexclusive GPL prior to the
present SCO's purchase (transfer of ownership)
from Novell is eliminated by sec 205(e) from the
protection of the GPL under the Copyright Act.
1. The GPL is not a transfer of copyright ownership, which is what this law refers to. It is a licence. Licences do not have to be signed, otherwise you'd need Bill Gates' signature in order to run a copy of Windows.
2. WTF does this have to do with Jon Johanssen?
But it may only last untill EUCD is implemented...
The UK EUCD implementation grants specific requirements that mean you must be allowed access to an unencrypted copy of the data if you have a legitimate reason for needing it... unforuntately the procedure for enforcement of this is ridiculously complicated (you must complain to the home secretary, IIRC).
I'm not sure how it'll work in norway, but I'd hope there would be some similar provision.
However face recognition might not be the perfectly viable solution since it has produced too many false positives in the past.
False positives aren't too much of a problem here. I think face recognition has about a 1/1000 false positive rate, which is a killer for crowd recognition, but would be entirely acceptable for this application (the result would be that 1 time in a thousand somebody tried to use someone else's passport, they wouldn't be caught).
False negatives are what we need to worry about in this situation... I don't know much about the statistics here. Anyone have any idea?
Except for the fact that nobody would accept your document as authoritative, because nobody has any authority to produce such a document, except perhaps the X Consortium, who have no interest in producing one, and would almost certainly be unable to agree with the details if there were.
There is generally speaking consisteny in Windows beause MS is accepted as having authority to declare how user interfaces should work there; the same is true of Apple having authority over OSX. Nobody has that authority over X, so it splinters into individual groups (GNOME, KDE, CDE, etc.) who have their own guidelines. So you then need to learn multiple guidelines in order to be able to use the same apps. And there are apps (e.g. emacs) which don't follow any of them.
I've always seen Forking as something of a blessing... it's the abandoned projects are the ones that are in danger.
Forking leads directly to abandoned projects as it dilutes the number of developers willing to work on each fork.
When I was working for a major global firm, and we dealt with small closed source development companies, we always had code escrow agreements. If the vendor went out of business or dropped support of the product, we had the ability to get the source and support it ourselves.
Now, I can assure you, you can only normally get this kind of agreement if you have a lot of spare cash to throw around. Throwing around IP rights like that limits a business that is being wound up from selling its products to another business, and that is something that smart companies will not do except for a good price.
I also wouldn't wager anybody's chances of making this happen with any kind of mass-marketed software, even if it is aimed at a specialised market. For instance: my company uses a piece of proprietary software to analyse web pages for search engine keyword placement as a method to attempt to improve rankings. I severely doubt that in any way could we persuade the company that wrote this software to agree to the sort of terms you are suggesting, unless we offered to pay them substantially more money than that software is worth to us.
And at the end of all this, what do you end up with? Well, pretty much the same as if you started using an open source product that stopped being developed.
that make differences in behaviour even more annoying - should I use Alt+F4 or Ctrl+W to close an application window, for instance?
I have never seen an application for which Alt+F4 doesn't work, and if there are any, it would be a violation of the Microsoft UI guidelines (Available here). Ctrl+W doesn't appear to actually be standardised, but as a de-facto standard often closes the document you're working with, rather than the entire application. Hope this helps!
This isn't how Linus has interpreted the GPL.
Maybe not. But if it ever came to court, it wouldn't be Linus that had to decide. And, by the straightest possible reading of the GPL, trying to divine the original intention of it when it was written (which is, I think, how a court would approach it), such things are probably allowed.
It all rests on what a derivitive work is.
A case to consider: if any software that depends on any other's internal workings in order to funtion correctly is a derivitive work of the other, then is my Windows software a derivitive of Windows, and if so, when will Microsoft be starting proceedings against me? My software "needs to know intimately" the Windows APIs otherwise it wouldn't work, and it "gets linked to [them] when it's loaded", using pretty much exactly the same mechanisms as a Linux kernel module.
Sorry, Tom, You've started chsshhhh chshhhshshhh -king up cshshhhhhhhhh...
Yeah, I hate it when things start cshing up too.
what if someone wrote another standard driver interface, separate from the kernel interface, wrote a device driver which implemented that and then wrote a GPL'd interface wrapper which translated the Linux module interface to that of the new standard?
Obviously, the wrapper interface code is now a derived work. However, does it also mean that because the new driver which uses a code interface which the GPL'd wrapper implements now is tainted by the GPL?
From the GPL:
If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it.
Therefore the GPL does not apply to the original devie driver unless and until it is distributed alongside the GPL'd interface wrapper linked together in a single compiled image (by my reading, although IANAL).
Also, does the driver become a derived work if the person who writes it initially does so to get some hardware working on his Linux box, rather than his other box which runs ObsureOS which also implements this standard device driver interface but the person hasn't installed the hardware on that machine yet?
I don't think that would make any difference: as long as the driver was useful without the GPL'd code it could "reasonably [be] considered [an] independent and separate [work] in [itself]"
to borrow the GPL's phrasing and adjust for the singular rather than the plural.
Dynamically linking to a GPL'd library does not _necessarily_ cause a program to become licensed under the GPL.
It has long been held in copyright law (through case history) that using an API does not, of itself, mean that the program that uses it is a derivitive of an implementation of that API.
This would be especially true if two implementations of the same API existed, which is probably the case in the nVidia drivers that were being discussed; there is a small driver module (whose source code is available) that merely provides an API layer between the kernel and another program (no source code available). My guess is that that program is simply a recompilation, with no source code changes, of the equivalent program under Windows, which would use a different compatibility layer.
It is therefore plainly not a derivitive of the Linux compatibility layer, because it could be used separately from it. And by the terms of the GPL, as it is a separable program that doesn't depend entirely upon the other, it does not have to be GPL licensed.
And the point of this would be...?
Hint:
1. The specifications probably change with each new card release
2. One of the parent posts gave an estimate of (IIRC) 1.8 Million lines of code in the library that they aren't giving out the source to. Are you volunteering to duplicate that?
How about: EBNF is the future of syntactic notation?