Why would I be compelled to transfer source code to anybody? The only thing that would require me to do so is the GPL. I don't have to agree to the terms of the GPL because I'm not making any copies of the software.
If the vendor picked the option of giving you the source code along with the binaries, they can be argued to both be part of one piece of software. You have to transfer the entirety of the software to avoid copyright triggering. Otherwise you are modifying, and that is a copyright "event". You cannot keep some of the software (the source code) and transfer the rest.
However, that is not the interesting case, as you are implying that the vendor picked the option of providing a written offer instead of source code.
They can't. The organisation I bought the machine from is only compelled by the GPL to provide source code to the person they distributed the binaries to, i.e. me.
This is not true. The GPL requires that the offer is made to "any third party".
I think you might have misread my comment. Why are you talking about redownloading the software to put it on a new machine? I am buying the machine with the software preinstalled. I'm not downloading or installing anything.
Because you can get around the GPL once that way. It is not really a hole worth being concerned about.
Does the person I sold the machine to have any right to ask me for source code? Why?
It gets tricky. Generally you are only allowed to pass on the entire software or none at all, not to modify it by only providing parts. Either the source code came with the software, in which case it could be argued that you have to transfer both binaries and source code in order to comply with copyright, or it came with a written offer. In the latter case the person you sold it to can just go directly to whoever made the written offer.
Notice that the recipient never actually has the right to ask for source code. They do not own the copyright, so they cannot enforce the license. All they can do is alert the copyright holder(s) of the license violation and hope that the copyright holder(s) care.
Even if you manage to circumvent the GPL this way, it is rather inefficient. You have to redownload the software every time you put it on a new machine, otherwise you're copying it. If you make scripts to do it for you, you are likely to end up modifying the software, which triggers copyright.
I missed that part, and yes, trying to cover it up only hurts. I still expect a fair number of management employees to walk away with the soundbite that GPL equals lawsuits.
Hopefully the management employees will also notice that the average number of GPL violation cases going to court is below 1 per year, and that most of the settlements are really, really cheap.
Hopefully they will not notice that there are very few developers of GPL'd software who are willing to defend it in court, and therefore the GPL can be ignored on most code if you are sufficiently brave.
It probably wouldn't have cost them as much as most likely it would have been settled out of court without the need for lawyers and court fees, the BSA just wants to get paid after all and will negotiate,whereas with the GPL there is NO negotiation nor compromise because like it or not that is the way RMS designed the license.
The vast majority of GPL violations get handled out of court. Anecdotal evidence seems to suggest that the payment in most cases is zero.
Most actual court cases around GPL software seem to be brought by Harald Welte, and he in particular settles almost all cases outside court.
Using code at all is unpredictably dangerous. In most cases, it is impossible for someone to prove that a particular piece of software does not incorporate any unlicensed third party code. Software patents make it all even murkier. Such is life with "intellectual property".
If you compile the software yourself from source, you have at least some chance at finding violations yourself. On the other hand, if you get handed a binary blob to redistribute, you better have a very trustworthy supplier.
More likely, the drives will release (assuming it ever makes to market) with a far lower rate (2x-4x BluRay speeds is my guess) to prevent excessive noise, heat and wear, and transfer speeds are never perfect so a full-disk copy will require about 10 hours.
You are ignoring the fact that density gains come not just from closer spacing of the tracks, but also from closer spacing of the data in each track (yes I know there is technically only one spiral track on an optical disc, but hopefully you understand what I mean.) If a BluRay is 50GB and the new discs are 300GB and the density increase is the same both along tracks and between tracks and rotational speed is kept the same, the new discs will have a 1x speed that is 2.4 times (square root 6) as fast as BluRay and it will take 2.4 times as long to read the entire disc.
The whole thing is pointless though, it is meant for professional use only which means they will build maybe a few thousand units in total. It will be unable to compete on price with hard drives or tape, so you only buy it if you need decade-long storage without maintenance (and good luck with that).
they could not show that any particular compiler or architecture made the predictions any better, just different.
If one of the architectures/compilers had come up with a constant "result = 1" or indeed any degradation of model performance at all, they would have been able to make much stronger statements in the abstract.
It is so unfortunate that academics do not have the wisdom of Slashdot available before they submit papers. Alas, that is the reality they have to live with.
When floating point roundoff errors grow big enough to affect the outcome of the simulation, you have long since reached the point where you are not predicting anything useful any longer.
This is not true. If the model predicts rain at 2 pm two days out and different rounding moves it to 3 pm, that is still a useful forecast in a lot of cases.
Enjoying decent performance. Doing weather forecasts slower than real time is a lot easier but somewhat less useful.
My interpretation of the abstract (I cannot access the actual paper) is that they could not show that any particular compiler or architecture made the predictions any better, just different. In that case you just go with whichever runs fastest.
How does that kind of acceleration affect total range? Valid concern with electrics, not so important with decently-powered IC engines.
It will not significantly affect range. You need to spend the same amount of energy to get to your desired speed whether you do it quickly or slowly, and the electric motor does not significantly gain or lose efficiency depending on load. The only real difference is the extra air resistance for being at the desired speed 4 seconds earlier.
Of course if you needlessly accelerate to a higher speed just to brake more at the next traffic light, you will lose range. The comparison is only fair if both cars actually achieve the desired speed.
Most wind turbines actually do not use neodymium magnets. They use plain old electromagnetic generators. However, to electromagnetic generators get inefficient at low speeds, so using electromagnetic generators also means using gears, possibly multi-stage. Coal, gas, and nuclear plants generally work with very hot steam. You can design your steam turbine for pretty much any rotational speed you want. No gears necessary.
So why not just increase the rotational speed of wind turbines? You lose aerodynamic efficiency when the tip speed gets to a reasonable fraction of the speed of sound. To avoid gears and use electromagnetic generators on a typical 3MW wind turbine, you would need the tips to go faster than the speed of sound. Extracting power from the wind while going at supersonic speeds is a yet unsolved problem.
If the use of permanent magnets was banned entirely from wind turbines, the market would not really change much. Some manufacturers would temporarily lose market share while they redesigned, but overall turbine price would not change dramatically.
So if I login to GMail with my phone and my desktop, if I log off on my desktop it should kill my phone too? How the hell is that better?
If you log in to GMail twice, you get two different cookies. In a sane world, when you hit "logout", the specific cookie gets invalidated and you have to log in again on that device if you want back in. Hotmail (seemingly unlike GMail) does not exist in a sane world.
That sounds unlikely. If you know where the signal is going to hop, it is trivial to follow. I have not heard of a standard that picks the next frequency in a cryptographically secure way, but I am prepared to be surprised of course.
Who said the headers have to be binary? Sure, they're compressed, but decompress them and they magically turn back in to plain text.
You can call all binary protocols "compressed" if you prefer. That does not make them less annoying to deal with. However, I think it is incorrect to call something compression when it is a domain-specific binary encoding.
So why is TCP/IP unworkable with binary framing but another protocol is not?
TCP/IP uses binary framing and is obviously not unworkable. If you mean ASCII framing, it would be unworkable to put <PROTOCOL>TCP</PROTOCOL><PORT>80</PORT> into every packet, the overhead would just be too great. (And yes, XML is probably the least space-efficient encoding in existence, but anything ASCII would be too verbose)
HTTP 2.0 handles its congestion control in a way that works with TCP. I am not sure what you mean by increasing the load on the network. Pure routers do not care if the packet is SCTP or QUIC or HTTP 2.0. Anything stateful hates dealing with lots of connections, so HTTP 2.0 wins there.
You do not want per-stream congestion handling for data of the same priority. One steady TCP connection is significantly better at hitting optimum bandwidth than multiple competing connections, particularly when the competing ones come and go rapidly.
Fair enough. Do you prefer reinventing TCP over UDP? That's called QUIC and might have a chance. Or an entirely different IP protocol? That's called SCTP and it is from 2000; it is used only by phone companies.
It annoys me that SCTP cannot get deployed, but it looks unlikely to change.
The "Oh shit I think I hit that guy so I'm GONE" reaction is where the driver went from "human who made a mistake" to "worthless piece of shit that deserves hard time in fed-max."
Driving away is a surprisingly common reaction, and in a lot of cases no link is found to being drunk or other "rational" reasons for doing so. I can understand why you advocate fed-max, but the fed-max prisons will be even more overcrowded if we somehow detect everyone who could do a hit-and-run and round them up.
Luckily most of us will never have to discover how we react in such a horrendous situation.
I'm firmly in the ASCII camp, but I think the draft itself lays out the reasons pretty well. They want to do TCP over TCP, and that requires chunking the data into frames with individual headers. If those headers have to be in binary, the whole thing is unworkable. Just like TCP/IP itself would be unworkable if the headers had to be cleartext.
How many consumer NAT routers allow SCTP through? How many enterprise firewalls?
Yes HTTP 2.0 is solving things at the wrong layer. Just like QUIC being UDP-encapsulated is completely backwards. However, such is life in the modern Internet. As IPv6 has shown, innovation on the basic protocol side is almost impossible, and anything new takes at least two decades to get minimal traction.
You cannot autoland all the time. Well you could, but the pilots would lose their ability to handle the aircraft by hand. It is almost certain that Asiana Flight 214 could have auto-landed, and that would have saved at least 2 lives. How many lives would it cost if pilots never did any flying except in emergencies?
One day that trade-off is probably going to swing towards not letting the pilots do anything useful...
Whats the use of preserving humanity if there are no humans around to "be human"
What is the use of doing anything which will only be useful after you are dead? Humans are sentimental creatures. Most people like to believe that something of value will survive, even if everyone dies.
End-of-the-universe theories are a bit depressing currently, but at least they get regular revision. Even black holes leave ways of figuring out what was lost inside them, in theory.
Re:Testing the character parsing of every web site
on
Fedora 19 Released
·
· Score: 1
More importantly, testing the character parsing and handling of both the installer and multiple other parts of the distribution. Whether it was a good idea to pick a challenging name is probably dependent on the observer.
Slashdot Mirror
Why would I be compelled to transfer source code to anybody? The only thing that would require me to do so is the GPL. I don't have to agree to the terms of the GPL because I'm not making any copies of the software.
If the vendor picked the option of giving you the source code along with the binaries, they can be argued to both be part of one piece of software. You have to transfer the entirety of the software to avoid copyright triggering. Otherwise you are modifying, and that is a copyright "event". You cannot keep some of the software (the source code) and transfer the rest.
However, that is not the interesting case, as you are implying that the vendor picked the option of providing a written offer instead of source code.
They can't. The organisation I bought the machine from is only compelled by the GPL to provide source code to the person they distributed the binaries to, i.e. me.
This is not true. The GPL requires that the offer is made to "any third party".
I think you might have misread my comment. Why are you talking about redownloading the software to put it on a new machine? I am buying the machine with the software preinstalled. I'm not downloading or installing anything.
Because you can get around the GPL once that way. It is not really a hole worth being concerned about.
Does the person I sold the machine to have any right to ask me for source code? Why?
It gets tricky. Generally you are only allowed to pass on the entire software or none at all, not to modify it by only providing parts. Either the source code came with the software, in which case it could be argued that you have to transfer both binaries and source code in order to comply with copyright, or it came with a written offer. In the latter case the person you sold it to can just go directly to whoever made the written offer.
Notice that the recipient never actually has the right to ask for source code. They do not own the copyright, so they cannot enforce the license. All they can do is alert the copyright holder(s) of the license violation and hope that the copyright holder(s) care.
Even if you manage to circumvent the GPL this way, it is rather inefficient. You have to redownload the software every time you put it on a new machine, otherwise you're copying it. If you make scripts to do it for you, you are likely to end up modifying the software, which triggers copyright.
I missed that part, and yes, trying to cover it up only hurts. I still expect a fair number of management employees to walk away with the soundbite that GPL equals lawsuits.
Hopefully the management employees will also notice that the average number of GPL violation cases going to court is below 1 per year, and that most of the settlements are really, really cheap.
Hopefully they will not notice that there are very few developers of GPL'd software who are willing to defend it in court, and therefore the GPL can be ignored on most code if you are sufficiently brave.
It probably wouldn't have cost them as much as most likely it would have been settled out of court without the need for lawyers and court fees, the BSA just wants to get paid after all and will negotiate,whereas with the GPL there is NO negotiation nor compromise because like it or not that is the way RMS designed the license.
The vast majority of GPL violations get handled out of court. Anecdotal evidence seems to suggest that the payment in most cases is zero.
Most actual court cases around GPL software seem to be brought by Harald Welte, and he in particular settles almost all cases outside court.
Using code at all is unpredictably dangerous. In most cases, it is impossible for someone to prove that a particular piece of software does not incorporate any unlicensed third party code. Software patents make it all even murkier. Such is life with "intellectual property".
If you compile the software yourself from source, you have at least some chance at finding violations yourself. On the other hand, if you get handed a binary blob to redistribute, you better have a very trustworthy supplier.
More likely, the drives will release (assuming it ever makes to market) with a far lower rate (2x-4x BluRay speeds is my guess) to prevent excessive noise, heat and wear, and transfer speeds are never perfect so a full-disk copy will require about 10 hours.
You are ignoring the fact that density gains come not just from closer spacing of the tracks, but also from closer spacing of the data in each track (yes I know there is technically only one spiral track on an optical disc, but hopefully you understand what I mean.) If a BluRay is 50GB and the new discs are 300GB and the density increase is the same both along tracks and between tracks and rotational speed is kept the same, the new discs will have a 1x speed that is 2.4 times (square root 6) as fast as BluRay and it will take 2.4 times as long to read the entire disc.
The whole thing is pointless though, it is meant for professional use only which means they will build maybe a few thousand units in total. It will be unable to compete on price with hard drives or tape, so you only buy it if you need decade-long storage without maintenance (and good luck with that).
You didn't read what I wrote.
To repeast myself:
they could not show that any particular compiler or architecture made the predictions any better, just different.
If one of the architectures/compilers had come up with a constant "result = 1" or indeed any degradation of model performance at all, they would have been able to make much stronger statements in the abstract.
It is so unfortunate that academics do not have the wisdom of Slashdot available before they submit papers. Alas, that is the reality they have to live with.
When floating point roundoff errors grow big enough to affect the outcome of the simulation, you have long since reached the point where you are not predicting anything useful any longer.
This is not true. If the model predicts rain at 2 pm two days out and different rounding moves it to 3 pm, that is still a useful forecast in a lot of cases.
WTF are these amateurs doing?
Enjoying decent performance. Doing weather forecasts slower than real time is a lot easier but somewhat less useful.
My interpretation of the abstract (I cannot access the actual paper) is that they could not show that any particular compiler or architecture made the predictions any better, just different. In that case you just go with whichever runs fastest.
How does that kind of acceleration affect total range? Valid concern with electrics, not so important with decently-powered IC engines.
It will not significantly affect range. You need to spend the same amount of energy to get to your desired speed whether you do it quickly or slowly, and the electric motor does not significantly gain or lose efficiency depending on load. The only real difference is the extra air resistance for being at the desired speed 4 seconds earlier.
Of course if you needlessly accelerate to a higher speed just to brake more at the next traffic light, you will lose range. The comparison is only fair if both cars actually achieve the desired speed.
Most wind turbines actually do not use neodymium magnets. They use plain old electromagnetic generators. However, to electromagnetic generators get inefficient at low speeds, so using electromagnetic generators also means using gears, possibly multi-stage. Coal, gas, and nuclear plants generally work with very hot steam. You can design your steam turbine for pretty much any rotational speed you want. No gears necessary.
So why not just increase the rotational speed of wind turbines? You lose aerodynamic efficiency when the tip speed gets to a reasonable fraction of the speed of sound. To avoid gears and use electromagnetic generators on a typical 3MW wind turbine, you would need the tips to go faster than the speed of sound. Extracting power from the wind while going at supersonic speeds is a yet unsolved problem.
If the use of permanent magnets was banned entirely from wind turbines, the market would not really change much. Some manufacturers would temporarily lose market share while they redesigned, but overall turbine price would not change dramatically.
So if I login to GMail with my phone and my desktop, if I log off on my desktop it should kill my phone too? How the hell is that better?
If you log in to GMail twice, you get two different cookies. In a sane world, when you hit "logout", the specific cookie gets invalidated and you have to log in again on that device if you want back in. Hotmail (seemingly unlike GMail) does not exist in a sane world.
That sounds unlikely. If you know where the signal is going to hop, it is trivial to follow. I have not heard of a standard that picks the next frequency in a cryptographically secure way, but I am prepared to be surprised of course.
Who said the headers have to be binary? Sure, they're compressed, but decompress them and they magically turn back in to plain text.
You can call all binary protocols "compressed" if you prefer. That does not make them less annoying to deal with. However, I think it is incorrect to call something compression when it is a domain-specific binary encoding.
So why is TCP/IP unworkable with binary framing but another protocol is not?
TCP/IP uses binary framing and is obviously not unworkable. If you mean ASCII framing, it would be unworkable to put <PROTOCOL>TCP</PROTOCOL><PORT>80</PORT> into every packet, the overhead would just be too great. (And yes, XML is probably the least space-efficient encoding in existence, but anything ASCII would be too verbose)
HTTP 2.0 handles its congestion control in a way that works with TCP. I am not sure what you mean by increasing the load on the network. Pure routers do not care if the packet is SCTP or QUIC or HTTP 2.0. Anything stateful hates dealing with lots of connections, so HTTP 2.0 wins there.
You do not want per-stream congestion handling for data of the same priority. One steady TCP connection is significantly better at hitting optimum bandwidth than multiple competing connections, particularly when the competing ones come and go rapidly.
The original Pacman has an integer overflow. AFAIK it cannot be exploited except for DoS, but still...
Fair enough. Do you prefer reinventing TCP over UDP? That's called QUIC and might have a chance. Or an entirely different IP protocol? That's called SCTP and it is from 2000; it is used only by phone companies.
It annoys me that SCTP cannot get deployed, but it looks unlikely to change.
The "Oh shit I think I hit that guy so I'm GONE" reaction is where the driver went from "human who made a mistake" to "worthless piece of shit that deserves hard time in fed-max."
Driving away is a surprisingly common reaction, and in a lot of cases no link is found to being drunk or other "rational" reasons for doing so. I can understand why you advocate fed-max, but the fed-max prisons will be even more overcrowded if we somehow detect everyone who could do a hit-and-run and round them up.
Luckily most of us will never have to discover how we react in such a horrendous situation.
I'm firmly in the ASCII camp, but I think the draft itself lays out the reasons pretty well. They want to do TCP over TCP, and that requires chunking the data into frames with individual headers. If those headers have to be in binary, the whole thing is unworkable. Just like TCP/IP itself would be unworkable if the headers had to be cleartext.
How many consumer NAT routers allow SCTP through? How many enterprise firewalls?
Yes HTTP 2.0 is solving things at the wrong layer. Just like QUIC being UDP-encapsulated is completely backwards. However, such is life in the modern Internet. As IPv6 has shown, innovation on the basic protocol side is almost impossible, and anything new takes at least two decades to get minimal traction.
You cannot autoland all the time. Well you could, but the pilots would lose their ability to handle the aircraft by hand. It is almost certain that Asiana Flight 214 could have auto-landed, and that would have saved at least 2 lives. How many lives would it cost if pilots never did any flying except in emergencies?
One day that trade-off is probably going to swing towards not letting the pilots do anything useful...
Most passenger jets can (category 3) auto-land these days. It is frequently used.
Whats the use of preserving humanity if there are no humans around to "be human"
What is the use of doing anything which will only be useful after you are dead? Humans are sentimental creatures. Most people like to believe that something of value will survive, even if everyone dies.
End-of-the-universe theories are a bit depressing currently, but at least they get regular revision. Even black holes leave ways of figuring out what was lost inside them, in theory.
More importantly, testing the character parsing and handling of both the installer and multiple other parts of the distribution. Whether it was a good idea to pick a challenging name is probably dependent on the observer.