For a more complicated example then, let A,B,C be large (say 10000*10000) sparse matrices and work out det(A,B,C). Letting decisions as to how to implement this be left to the compiler and library authors is a good idea. (the standard library won't know the various reduction rules, or even that det(AB)=det(A)det(B) for square matrices...) John
where A,B,C,D are n*n matrices for which a typical compiler (even modern ones!) will compute A+B, store it in an intermediate array, D', say, and then compute D=D'+C. Doing D=A+B+C is quicker, but requires the knowledge that (mathematical) arrays are being dealt with -- this is the benefit of abstraction. and consider
x=det(A*B*C)
how many otimisers will notice that you can evaluate the matrix determinants and then multiply? (this is especially useful when the matrices are sparse). The morale: what a compiler doesn't know, it cannot be expected to figure out by brute computational force (that's why humans still do the research:-) ). John
Turing completeness does NOT imply that languages are equally powerful. Any program written for, say, turing complete language A, may be emulated by a program written in turing complete language B, Given sufficient memory and time in which to perform the computation. Similarly for the other way round.
If you want an example, try writing a decent fast fourier transform, or a 3d image manupulation system in TeX (which alas is also turing complete)
in short, turing completeness implies that the language is equivalent if you have a potentially infinitely large memory and a potentially infinitely fast processor John
You should take note of how recently the number zero and the equals (and plus etc.) symbols were invented, and what the Greek/Arabic mathematicians did without them. John
but mp3 encoders (like bladeenc) are opensource. if they stumble on the watermark, we can rewrite the encoding logic.
You (or at least someone) would still have to figure out HOW to rewrite the encoding logic.
also, if the watermark 'steals' lower order bits, its quite easy to rewrite those bits with our own. I'll certainly trade a bit of my own white-noise over their stupid fingerprints!
Expect them to be more clever than that -- there are literally trillions of ways to represent a 1second 16bit 2channel digital sound wave such that the difference is not noticeable -- even more if a little distortion is allowed. The error-inducing method the previous poster referred to was way of adding small errors that were AMPLIFIED by the psycho-acoustic models used in MP3 compression.
all you have to do is put the signal (digitally) thru a filter of some sort - like a highpass peaking filter and voila; the lower order bits (and all of them, actually) become changed. GONE is their fingerprint (and good riddance). and you have the benefit of having a slightly more 'present' upper end (which is great for the car).
as before -- using lower order bits alone is primative (and better methods are known -- have you ever, for example, looked at the various methods of steganographic encoding???) John
It depends upon who controls the use of the CDDA mark -- though clearly one of the two statements is false. (Possibly common law could come into this, since it is accepted that CDDA's play in CD players and adhere to a know format).
Does anybody know about the legal aspects in a useful level of detail?? John
If I can hear it, i can record an approximate likeness of it -- you forget about quality
If I can see it, I can record it -- same problem (i.e. VHS doesn't record broadcast quality TV, if you're hallucinating you can't record that either...)
In particular, with reference to the 'If I can see, I can record' -- if you grab a DVD through a graphics driver intercept, wouldn't you have to recompress the data (and in the meantime, where do you put those gigs of video?) John
Finally, RE: junking X altogether instead of extending the X prototcol... I'm in favor of keeping X and just extending the protocol. We have such a large application base locked up in X that to toss X would be to throw the baby out with the bathwater. X is crufty, and could use a core protocol update, but it's also still good enough for most every day work. We need to either update the protocol (yeah right, like the Open Group is going to bother), or extend it server side.
The problem with X has parallels with those of the FORTRAN language -- people are producing various requests for- and implementations of- extensions for the language -- but anything that goes into the standard has to be implemented separately by EVERYBODY.
Extending X has similar problems, since there is no standard way by which an arbitrary X server can be extended.
To quote the usual quote -- Keep It Simple Stupid -- X already has FAR too much baggage, and adding more isn't going to help it much in the short run (since it takes too long to produce a good implementation), nor in the long run (X's complication gives no real advantages -- network transparency is far better implemented with an extensible window system and network transparent IPC).
In short, the feature set of a good display architecture needs to be chosen VERY CAREFULLY, not hacked on and extended at will by any old person who comes along. (Consider the question of when X is going to drop its outdated and outmoded font system, or its poorly inadequate pixel based rendering system) John
Forcing the structures in the source code, by (for example) inventing a new language that has uses XML or SGML for the 'authoratative' version sounds like what you are talking about -- this allows for some of the fancy editing features in Visual Braindamage to be added trivially to any such language.
Question: What are pros/cons of what I shall term 'enforced coding structures' -- is there a proper term for this? John
In simple terms, the autoconf-automake-libtool chain needs to be unified into a SINGLE program with a SINGLE database of bits and pieces (which could possibly be overlaid by a user's own database). First, there is the speed issue. Second, the problem that version inconsistencies produce (how may times have you had automake or authconf complain...). Third, and most importantly, each part of the toolchain mentioned is a single-point-of-failure in the build process -- reducing that to number to two-ish (gcc,NewMakeTool,possibly libraries) would make building a lot easier, and predictable (the predictability is NECESSARY if you want to distribute source and expect it to compile the same way on someone else's box as it does on yours) John
You ARE allowed to charge for the software. The only thing is that the person to whom you sell GPL'd software has as much of a right to sell it, once he/she has it, as you have. John
very true -- most people would be happy to spend $5 for a TV quality version rather than $30 for the DVD quality version (which can then only be viewed at TV quality anyhow -- the lost detail won't be missed) John
All you need is a pressing plant in the east somewhere (I recall that an anti-piracy organisation in China was withdrawn after contracts were taken out against its management by the various large-scale copying operations over there) John
Just to make the obvious point -- it will be some time before the average consumer can download a multi-gigabyte movie and blow it to their own media (a high speed connection is required). John
In some sense, (take this whatever way you like), Caldera was a 'sue Microsoft' spinoff from Novell, with Ray Noorda (ex. Novell head honcho who had a vendetta/blood-hunt to pursue) at the Helm. John
If I want, say, small (i.e. 2-3 entry) submenus to be expanded inline (like in Office 2000), or ANY window's menubar to be at the top of the screen, or ANY window's toolbars to hide when I lift my hand off the mouse (given one of the M$ mice), then how do I configure the Window Manager to do this?
Note that it has nothing to do with content -- with the menu example, any submenu that has 2 entries will be expanded inline.
The problem, which stems from X, is that there is no enforced abstract structure to the windows -- they are all disconnected rectangles. John
Firstly, its not about being monitored -- the car monitors the speed limit and enforces it.
A manual override would be useful, but it should be such that only the police could place it back on, and a car with the manual override engaged SHOULD be able to be tracked (since an emergency situation is the only place where it should be used -- and in any case, you should have to go to a police station and justify your actions)
John
Re:Redundant, Concurrent Efforts are Inefficent
on
The ROX Desktop
·
· Score: 1
Not when the effort required to organise one unified effort outweighs the effort to have multiple independant ones, as is the case in the OSS community. John
Slashdot Mirror
For a more complicated example then, let A,B,C be large (say 10000*10000) sparse matrices and work out det(A,B,C). Letting decisions as to how to implement this be left to the compiler and library authors is a good idea. (the standard library won't know the various reduction rules, or even that det(AB)=det(A)det(B) for square matrices...)
John
John
Why don't you spell Europeanz with a z while you're at it -- it makes the same sound :-)
John
Turing completeness does NOT imply that languages are equally powerful. Any program written for, say, turing complete language A, may be emulated by a program written in turing complete language B, Given sufficient memory and time in which to perform the computation. Similarly for the other way round.
If you want an example, try writing a decent fast fourier transform, or a 3d image manupulation system in TeX (which alas is also turing complete)
in short, turing completeness implies that the language is equivalent if you have a potentially infinitely large memory and a potentially infinitely fast processor
John
You should take note of how recently the number zero and the equals (and plus etc.) symbols were invented, and what the Greek/Arabic mathematicians did without them.
John
John
It depends upon who controls the use of the CDDA mark -- though clearly one of the two statements is false. (Possibly common law could come into this, since it is accepted that CDDA's play in CD players and adhere to a know format).
Does anybody know about the legal aspects in a useful level of detail??
John
John
For the loss of quality, you may as well buy it on cassette and save a little money.
John
Extending X has similar problems, since there is no standard way by which an arbitrary X server can be extended.
To quote the usual quote -- Keep It Simple Stupid -- X already has FAR too much baggage, and adding more isn't going to help it much in the short run (since it takes too long to produce a good implementation), nor in the long run (X's complication gives no real advantages -- network transparency is far better implemented with an extensible window system and network transparent IPC).
In short, the feature set of a good display architecture needs to be chosen VERY CAREFULLY, not hacked on and extended at will by any old person who comes along. (Consider the question of when X is going to drop its outdated and outmoded font system, or its poorly inadequate pixel based rendering system)
John
Linus said that he didn't want as long a development cycle as last time...
John
John
But picoJava is hardwired Java -- the point about Crusoe is that basically any VM can be soft/hardwired.
John
Forcing the structures in the source code, by (for example) inventing a new language that has uses XML or SGML for the 'authoratative' version sounds like what you are talking about -- this allows for some of the fancy editing features in Visual Braindamage to be added trivially to any such language.
Question: What are pros/cons of what I shall term 'enforced coding structures' -- is there a proper term for this?
John
In simple terms, the autoconf-automake-libtool chain needs to be unified into a SINGLE program with a SINGLE database of bits and pieces (which could possibly be overlaid by a user's own database). First, there is the speed issue. Second, the problem that version inconsistencies produce (how may times have you had automake or authconf complain...). Third, and most importantly, each part of the toolchain mentioned is a single-point-of-failure in the build process -- reducing that to number to two-ish (gcc,NewMakeTool,possibly libraries) would make building a lot easier, and predictable (the predictability is NECESSARY if you want to distribute source and expect it to compile the same way on someone else's box as it does on yours)
John
You ARE allowed to charge for the software. The only thing is that the person to whom you sell GPL'd software has as much of a right to sell it, once he/she has it, as you have.
John
very true -- most people would be happy to spend $5 for a TV quality version rather than $30 for the DVD quality version (which can then only be viewed at TV quality anyhow -- the lost detail won't be missed)
John
All you need is a pressing plant in the east somewhere (I recall that an anti-piracy organisation in China was withdrawn after contracts were taken out against its management by the various large-scale copying operations over there)
John
Just to make the obvious point -- it will be some time before the average consumer can download a multi-gigabyte movie and blow it to their own media (a high speed connection is required).
John
Redhat 5.0->6.0 changed the kernel from 2.0 to 2.2. That's pretty major
John
In some sense, (take this whatever way you like), Caldera was a 'sue Microsoft' spinoff from Novell, with Ray Noorda (ex. Novell head honcho who had a vendetta/blood-hunt to pursue) at the Helm.
John
If I want, say, small (i.e. 2-3 entry) submenus to be expanded inline (like in Office 2000), or ANY window's menubar to be at the top of the screen, or ANY window's toolbars to hide when I lift my hand off the mouse (given one of the M$ mice), then how do I configure the Window Manager to do this?
Note that it has nothing to do with content -- with the menu example, any submenu that has 2 entries will be expanded inline.
The problem, which stems from X, is that there is no enforced abstract structure to the windows -- they are all disconnected rectangles.
John
- Get a limilited liability company
- Get a loan
- Sue your company (and pocket the cash)
- Bankrupt your company
pop quiz -- what's wrong with this thinking?John
Firstly, its not about being monitored -- the car monitors the speed limit and enforces it.
A manual override would be useful, but it should be such that only the police could place it back on, and a car with the manual override engaged SHOULD be able to be tracked (since an emergency situation is the only place where it should be used -- and in any case, you should have to go to a police station and justify your actions)
John
Not when the effort required to organise one unified effort outweighs the effort to have multiple independant ones, as is the case in the OSS community.
John