Yeah - if you do the decryption in hardware it's
apparently no problem. Also not all DVDs use
encryption, so you can play these anyway with opensource players.
I've been using Linux since kernel version 0.99pl14...
I love Linux, and yes Linux is a nightmare to install. Many things have improved, that's certainly true, but: I recently added an IDE CD-burner to my PC.
I needed to: recompile the kernel, setup the SCSI emulator, find burner software... I spend half a day on this. I want Linux to improve - in most cases it's much better than Windows already. I think pretending that it is in all cases is counter-productive.
Isn't Moody even more wrong, then the article
points out? Any given user of Linux can not
experience the aggregate of security bugs -
in the "worst" case he'd have the distribution
with the most reported bugs. In this case RedHat
with 38.
This of course working with *his* assumption that
the number of reported bugs is a measure of
security - which it is clearly not, as has been
pointed out by many people here already.
Guess it's pointless to reply to it now, but here it goes:
CPUs - they aren't built for a specific speed
Correct, but that's true for basically any chip. The issue is that they are tested for a specific speed.
push it until it's not stable
And that is exactly what you can not do.
The problem is that the point it will stop working correctly depends on the task the processor is executing. There is a different max speed for multiplying 3FF * 45C than for 803 * 3DD for example.
If you don't know the critical path of a design you can't be sure it will work at a certain speed - you can only make a guess.
When some company manufactures their chip they'll test it with the right patterns - they rate it for the highest speed grade at which it can pass the test.
If you get a 600 MHz rated devices, and it's possible to buy a 650 MHz device - well duh it means your chip has been tested for 650 MHz and failed. That's why they sell it cheaper...
In order for a flipflop to switch correctly you need to maintain a setup time (data valid before clock edge) and a hold time (data valid after clock edge). The actually available setup time is determined by the time it takes data to propagate through the logic in front of the flipflop.
Lets take a simple setup (an AND gate with input and output flipflops).
This needs to be bigger than the required input setup time.
Also you need to find the critical path for this analysis - the wiring delay coming from the upper flipflop could easily be 10 times longer than from the lower flipflop. Relevant obviously is the max.
Now the problem with measuring whether the circuit works reliably is that depending on the data the critical path contributes to the result or not. In the above example, if the lower ff has the longer wiring delay associated, this will only show up if the upper ff has a '1' as an output. (because 0 AND x = 0; 1 AND x = x)
You can not realiable test the device without having the designer's test patterns. You can't calculate these patterns without knowing the exact structure of the device.
If you overclock you ought to be aware that you do not and can not know exactly what you are doing.
Well my impression was that this is not the whole problem. But as to the problem described on the web site: I don't think it's real.
Sure the GPL may contain restrictions the author doesn't actually want. This is only relevant if that license is the entire basis of your contract with the author. Contracts in general do not have to be in written form. Neither do amendments to contracts.
E.g. waving to a waitress with two fingers can create a contract: "I want two beers" (well assuming the right environment).
Also drinking a beer which has been put in front of you creates a contract - now you have agreed to pay for it. (This is true even if you didn't actually order the beer.)
The same thing applies with KDE. The programs calls the QT library. That means the author calls for that library to be included. So that constitutes permission to link against it. That's all that's needed.
It gets more interesting if the KDE code contains other GPL code. The example I've been told about is kghostscript, I assume there is more. (Is there a list anywhere?)
Then if you feel that QPL+GPL doesn't work (I'm not really clear on this) you'd have to get the author of ghostscript to agree to the linking. Or politely ask the Troll people to make a change to their license.
However I don't see that changing the KDE license would: - be necessary in any way or - provide any benefit at all
The internet was never intended for kids. In the beginning of the net it was almost exclussively accessible by adults, and there was always talk about sex.
Nobody ever claimed - "here is the internet, it's a free babysitter." Protecting children from the net is easy: don't give them access. You need a babysitter? Hire a babysitter!
I think there is a mistake in your post: in Europe the VAT does not cascade - it is only charged on the sale to the end consumer. E.g. a business to business sale is not subject to VAT. This at least applies to Germany, but I don't think it's handled differently in other European countries.
As usual this article mentions a lot of compatibiliy issues. This is fine execept: It doesn't say what they are!
If you really want something fixed, it would make a lot of sense to state what that is.:^)
As for the distribution being illegal... it's a somewhat far fetched argument. Remember the KDE is stuff is released by KDE under GPL license. If something doesn't make sense in the way they license the code - who exactly could sue? Seems like only KDE developers could do that.
That probably is the reason why most distributions don't care - it's a thoroughly mute point.
It would really only matter if KDE would break someone else's copyright - they are perfectly entiteled to break their own.
If they violate someone else's rights: please explain! It really doesn't make sense to write an article and assume everybody knows every detail of the discussed topic...
Well the *author* of the code can attach any restrictions to the license of the code. So you can say - "I wrote this, you can use it under the GPL, but...".
It's different if you modify someone else's code which is under the GPL: then you are not licensed to place additional restrictions on it.
I think it is about censorship, alright. The media does not need to rely on legally obtained information, only. Let's say some employee provides information about his tobacoo-selling company: they don't tell the truth about the harmful effects of their product. The employee might well break an NDA-like agreement doing so - but the newspaper publishing his info does not. More: they are not required to name their sources, even if the sources are criminal, even if the information was illegally obtained. These are rights any print-journalist can claim, just because of his profession. True there is no government action but: a constitutionally protected article has been removed from a website, as a result of Adobe making threads. I can't read that article anymore - it has been censored. And that is the very essence of censorship: denying access to information.
I think it's good that Opera is making progress, mainly because Netscape is so unstable. Anything (Mozilla, Konqueror or Opera) which could replace it, would be great for me.
However looking at the progress the various programs make, wouldn't konqueror be the best choice for Linux? It's license seems to be the "most free" of the web browser with the required features. (Not that it is sufficiently stable for daily use, either.)
Yes electrons travel slower, but there is an additional problem: On a computer chip the transfer of information is slowed down further by the line capacitance and line resistance. The capacitance is proportional to the distance of the wires to each other and to the length of the wires. The resistance is inversely proportional to the width of the wire.
The time constant is determined by R*C. So all three factors slow down signal propagation as you move to smaller geometries/higher density on the chip.
As for transistors - they can switch pretty fast anyway - the delay is practically all routing.
The trick of course is to achieve high integration - you need to be able to manufacture highly integrated optic.
Interesting yes, but I believe it when I see it. (Fair enough considering that it's optic.:)
It's pretty tough to *prove* that something can not be done. In a few cases it's possible, but every now and then you find that a problem can almost be solved or that you can get close enough
to the solution for all practical purposes.
E.g. by using heuristical algorithms.
The problem with postulating "it can't be done" is
that sometimes useful areas of research get blocked - this happened with neural nets for a long time. It was postulated that these nets could
never implement an XOR function. This is only true for nets which lack an intermediate layer.
Of course, today neural nets can implement an XOR...
So it's easy to miss something when you declare something as unsolvable. Even when you have a mathematical proof (as in the neural net example).
It follows that "can't be done" should only be used with the utmost care - a whole book of "unsolvables" seems ludicrous in this respect - unless he's merely re-iterating what everybody knows already.
Slashdot Mirror
Yeah - if you do the decryption in hardware it's
apparently no problem. Also not all DVDs use
encryption, so you can play these anyway with opensource players.
I love Linux, and yes Linux is a nightmare to install. Many things have improved, that's certainly true, but: I recently added an IDE CD-burner to my PC.
I needed to: recompile the kernel, setup the SCSI emulator, find burner software... I spend half a day on this. I want Linux to improve - in most cases it's much better than Windows already. I think pretending that it is in all cases is counter-productive.
Isn't Moody even more wrong, then the article
points out? Any given user of Linux can not
experience the aggregate of security bugs -
in the "worst" case he'd have the distribution
with the most reported bugs. In this case RedHat
with 38.
This of course working with *his* assumption that
the number of reported bugs is a measure of
security - which it is clearly not, as has been
pointed out by many people here already.
CPUs - they aren't built for a specific speed
Correct, but that's true for basically any chip. The issue is that they are tested for a specific speed.
push it until it's not stable
And that is exactly what you can not do.
The problem is that the point it will stop working correctly depends on the task the processor is executing. There is a different max speed for multiplying 3FF * 45C than for 803 * 3DD for example.
If you don't know the critical path of a design you can't be sure it will work at a certain speed - you can only make a guess.
When some company manufactures their chip they'll test it with the right patterns - they rate it for the highest speed grade at which it can pass the test.
If you get a 600 MHz rated devices, and it's possible to buy a 650 MHz device - well duh it means your chip has been tested for 650 MHz and failed. That's why they sell it cheaper...
Hm...
In order for a flipflop to switch correctly
you need to maintain a setup time (data valid
before clock edge) and a hold time (data valid
after clock edge). The actually available setup
time is determined by the time it takes data to
propagate through the logic in front of the flipflop.
Lets take a simple setup (an AND gate with
input and output flipflops).
FF ---
AND -- FF -- Y
FF ---
(Wish I could use a decent ASCII graphic here...)
Now the actual setup time on flipflop "y" is:
clock_period - (output_delay + wiring_delay + logic_delay)
This needs to be bigger than the required input
setup time.
Also you need to find the critical path for this
analysis - the wiring delay coming from the upper
flipflop could easily be 10 times longer than
from the lower flipflop. Relevant obviously is
the max.
Now the problem with measuring whether the circuit
works reliably is that depending on the data the
critical path contributes to the result or not.
In the above example, if the lower ff has the
longer wiring delay associated, this will only
show up if the upper ff has a '1' as an output.
(because 0 AND x = 0; 1 AND x = x)
You can not realiable test the device without
having the designer's test patterns. You can't
calculate these patterns without knowing the
exact structure of the device.
If you overclock you ought to be aware that you
do not and can not know exactly what you are doing.
Well my impression was that this is not the
whole problem. But as to the problem described
on the web site: I don't think it's real.
Sure the GPL may contain restrictions the author
doesn't actually want. This is only relevant if
that license is the entire basis of your contract
with the author. Contracts in general do not
have to be in written form. Neither do amendments
to contracts.
E.g. waving to a waitress with two fingers
can create a contract: "I want two beers" (well
assuming the right environment).
Also drinking a beer which has been put in front
of you creates a contract - now you have agreed
to pay for it. (This is true even if you didn't
actually order the beer.)
The same thing applies with KDE. The programs
calls the QT library. That means the author calls
for that library to be included. So that
constitutes permission to link against it.
That's all that's needed.
It gets more interesting if the KDE code contains
other GPL code. The example I've been told
about is kghostscript, I assume there is more.
(Is there a list anywhere?)
Then if you feel that QPL+GPL doesn't work
(I'm not really clear on this) you'd have to
get the author of ghostscript to agree to the
linking. Or politely ask the Troll people to
make a change to their license.
However I don't see that changing the KDE license would:
- be necessary in any way or
- provide any benefit at all
The internet was never intended for kids.
In the beginning of the net it was almost
exclussively accessible by adults, and there
was always talk about sex.
Nobody ever claimed - "here is the internet, it's
a free babysitter." Protecting children from the
net is easy: don't give them access.
You need a babysitter? Hire a babysitter!
I think there is a mistake in your post:
in Europe the VAT does not cascade - it is only
charged on the sale to the end consumer. E.g.
a business to business sale is not subject to
VAT. This at least applies to Germany, but I don't
think it's handled differently in other European
countries.
I mean look at the yellow stone picture - :^)
bare rock, stiff pine trees pointing into
the sky, the wetness of the river washing the
rocks off.
BTW: how come wired didn't provide examples of
the images which passed the filter?
As usual this article mentions a lot of compatibiliy issues. This is fine execept:
:^)
It doesn't say what they are!
If you really want something fixed, it would
make a lot of sense to state what that is.
As for the distribution being illegal... it's
a somewhat far fetched argument. Remember the
KDE is stuff is released by KDE under GPL license.
If something doesn't make sense in the way they
license the code - who exactly could sue?
Seems like only KDE developers could do that.
That probably is the reason why most distributions
don't care - it's a thoroughly mute point.
It would really only matter if KDE would break
someone else's copyright - they are perfectly
entiteled to break their own.
If they violate someone else's rights: please explain!
It really doesn't make sense to write an article
and assume everybody knows every detail of the
discussed topic...
Well the *author* of the code can attach any ...".
restrictions to the license of the code.
So you can say - "I wrote this, you can use it
under the GPL, but
It's different if you modify someone else's code
which is under the GPL: then you are not licensed
to place additional restrictions on it.
Gee - doesn't anyone remember the Pamela Anderson
video tapes? It doesn't matter for freedom of
press that information was obtained illegaly.
I think it is about censorship, alright. The media does not need to rely on legally obtained information, only. Let's say some employee provides information about his tobacoo-selling company: they don't tell the truth about the harmful effects of their product. The employee might well break an NDA-like agreement doing so - but the newspaper publishing his info does not. More: they are not required to name their sources, even if the sources are criminal, even if the information was illegally obtained. These are rights any print-journalist can claim, just because of his profession. True there is no government action but: a constitutionally protected article has been removed from a website, as a result of Adobe making threads. I can't read that article anymore - it has been censored. And that is the very essence of censorship: denying access to information.
I think it's good that Opera is making progress,
mainly because Netscape is so unstable.
Anything (Mozilla, Konqueror or Opera) which could
replace it, would be great for me.
However looking at the progress the various
programs make, wouldn't konqueror be the best
choice for Linux? It's license seems to be the
"most free" of the web browser with the required
features. (Not that it is sufficiently stable
for daily use, either.)
Well I run KDE on my desktop - but that doesn't
:^)
stop me to run Gnome applications.
For example Gaim is a great program, so I use
that.
I'd be interested to see a comparison between
KDE/Gnome apps - which has the better application
for which task?
Then I want to pick and choose.
Don't forget that PGP signing and ecrypting are totally independent - just because you encrypted it doesn't mean you have to sign it!
Yes electrons travel slower, but there is an
:)
additional problem: On a computer chip the
transfer of information is slowed down further
by the line capacitance and line resistance.
The capacitance is proportional to the distance
of the wires to each other and to the length of
the wires. The resistance is inversely
proportional to the width of the wire.
The time constant is determined by R*C. So
all three factors slow down signal propagation
as you move to smaller geometries/higher density
on the chip.
As for transistors - they can switch pretty fast
anyway - the delay is practically all routing.
The trick of course is to achieve high
integration - you need to be able to manufacture
highly integrated optic.
Interesting yes, but I believe it when I see it.
(Fair enough considering that it's optic.
The problem with postulating "it can't be done" is that sometimes useful areas of research get blocked - this happened with neural nets for a long time. It was postulated that these nets could never implement an XOR function. This is only true for nets which lack an intermediate layer.
Of course, today neural nets can implement an XOR...
So it's easy to miss something when you declare something as unsolvable. Even when you have a mathematical proof (as in the neural net example).
It follows that "can't be done" should only be used with the utmost care - a whole book of "unsolvables" seems ludicrous in this respect - unless he's merely re-iterating what everybody knows already.