Further proof that no one cares: Steam's Hardware Survey March 2010. Most prevalent resolution amongst gamers? 1280x1024, at 19%. Second place is 1680x1050, at 18%. Neither of those are particularly high, with the highest resolution in the survey being 1920x1200 at 6% and "Other" is only 3.4%.
His point, and I have to agree with it, is that I had 1280x1024 in 1995 and 1600x1200 a few years later. And incidentally the displays were brighter and had a larger color gamet.
Besides when his eyes go in a few years he won't care about the high resolutions anymore.
Because he is staring at low resolution screens! For the sake of your eyes, get a high resolution display!
Re:Read it! That was taken way out of context.
on
Google's Evil NDA
·
· Score: 1
I read it again and think that mentioning that you are interviewing there is no problem at all. That you are you know before you sign the NDA, and ANYTHING you know before you sign the NDA is explictly NOT covered under the NDA.
Read it! That was taken way out of context.
on
Google's Evil NDA
·
· Score: 5, Informative
What you are not allowed to do in the section mentioned there is use the fact that you had a meeting with google in your marketing or press releases. It does not say you can never mention google, as the summary indicates. For individuals, the most likely and probably only implication of the section is that you might not be able to use an employment offer from google as a bargening chip with other companies, but this isn't clear to me, as I don't know if that can be considered in the classes of things you cannot disclose from other sections (or if by nature a job offer to you does not ential a duty of confidentiality on your part since it isn't related to trade secrets or anything else explicitly mentioned).
I'm as skeptical as the next nerd, but it's still essential for respected scientists to conduct the tests, do the math, and come up with an answer, even if just to debunk it formally. Anything less is a negation of what science is supposed to be about, and reduces scientists to the level of the Holy Office of the Inquisition, condemning a theory without testing it.
No. They want someone to take it seriously, patent the device, hire some people to perform the tests, and try to publish the results. When they submit their paper they will get an instant "jury" to tell them what additional tests they should do and where they were anything less than convincing. It's called peer review. This is just publicity.
It isn't up to the "respected scientists" to refute this, it is up to them (the inventors) to produce enough evidence that their device cannot simply be blown off as a crackpot perpetual motion machine.
Besides, I don't see any "theory" in their published information. So what exactly are the "respected scientists" suppose to test? Anything less than bothering to write up what you did before trying to convince people you are right is worse than a negation of what science is about, it is just lazy and insulting (on the off chance this isn't a crackpot thing).
Everything they ask for would be handled by the normal peer review of a paper submission. It might be a fun paper to review too (some are much more interesting than others, as I am sure you can imagine).
the biggest obstacle is public perception of anything with "nuclear" in the name
And that is clearly why certain members of the Bush administration (namely Bush) call it nucular instead of nuclear. Opinion polling found this to be a more friendly name.
A while ago I read the RFC. It is very scary. Multihoming as proposed moves things like name resolution into the kernel.
I will grant SCTP does some neet stuff, the best is that it allows independent non-mutually-blocking streams over one connection. It also has state cookies, yum.
SCTP tries to be all things to all people in one protocol. It reads as though they just decided the whole layered protocol thing was overrated and shoved every new feature into this one layer.
On network drivers at least, the driver has such a small interface the driver is put through a full formal methods based proof system to prove the driver doesn't have certain classes of bugs. I hear MS has quite a cluster devoted to model checking network drivers. Granted this doesn't work for all classes, but at least some get fairly rigerous verification. I personally wish Linux drivers got this kind of checking.
No, this will not exist forever (without serious hacking of the physical box). Imagine a world were all interconnects (including to the speakers) are encrypted with a one time key, digital, etc etc, and no licensed box may have analog output. (Obviously, speakers will need to do D to A and have their own amp). Now stop imagining and just wait a few years.
So unless you want to put a mic in front of a speaker (how many op-amps, DAC, and ADC are involved in that path?), you are out of luck with. No line out for you!
Oh my. That some moderator modded the parent "Insightful" rather than 'Funny' says something about how far we are from "News for Nerds". I guess this is now "News for people who think computers can be kind of cool, sometimes".
For those of you still scratching your heads, go do some basic research on how kernels development use to work. Reading LWN.net archives might be a good start.
Ah I remember the old 1.2/1.3 days when 1.2 only supported one instance of my scsi cards, and 1.3.(hundred something) didn't like other hardware. Sorry I think 1.2.x (for some x I don't remember, but could look up:) ) was the oldest kernel I ever was doing anything useful with, so I guess I am not a true 'Nerd'.
Um, I don't think you know much about Alpha or NT for DEC Alpha. First off, all instructions on Alpha are 32 bits long, and all address 64 bit address spaces. There are no "64-bit instructions" if you mean instruction length and there are no "32-bit instructions" if you mean address mode. Of course when you say "32-bit instructions in parallel" I have no clue what you really mean.
Now, certain compilers and OSes tried to ease the transition by arranging that nothing an application would ever allocate would fall outside the first 32 bits of address space, thus applications that cast pointers to ints (totally broken as they can be different sizes and are on alpha) wouldn't brake due to the developer's incompetence.
If you mean FX32, then we are talking about a binary translator for x86 -> alpha. This allowed you to run x86 win32 binaries on alpha, translating the instructions at run time, and forwarding the library and system calls to the appropriate native alpha OS component (as the OS did not itself run under FX32, but was native alpha code).
Of course, NT for alpha was indeed a total hack. There was special PAL code that made the paging mechanism act more x86 like. PAL code to make exceptions behave the way NT wanted them to. There were plenty of very hackish things about that port, but the whole "n-bit instruction" thing you mention is just gibberish.
I speak this as someone who has run alphas for many years (I have a 21064 still running, amoungst others) and who has recently written an alpha backend to an optimizing compiler.
Yea, um, sure. Have you ever gone to one of those so called "classes" future MBA's have to take? Thought not. But perhaps you would get bored of learning how to make Excel make graphs. Oh, and think about what that "M" stands for, and realized that most of the torture refered to by grad students (and by this I mean phd students, not masters students) is most apparent after they prelim or qual (depending on the school). Before that, life isn't too much worse off than for a masters student.
>Saying Java isn't cool is like saying Scheme or ML > isn't cool. It's just a personal preference,
No, Scheme and LISP (as well as modern ML varients) have features that are more powerful than Java. They can express things that cannot be expressed in Java, mainly code generation and modification at runtime, closures (Java has an ugly hack added for AWT), continuations, syntax extention, etc.
It is said that everyone eventually settles on some language on the power spectrum. You look at the features of your language and are reasonably satisfied. You look at simpler/less powerful languages and you wonder "how could anyone program effectively without feature x" (where x is one of you favorite features not available in the simpler language). You look at more powerful languages, and being rather satisfied with your current choice, say "What do I need feature y for? I don't use it now and see no reason too since I can program what I need with my language." (where feature y is a feature the more powerful language has and its supports make a great deal of hoopla over, whereas your language does not).
I was happy with C++, I could do amazingly strange and terse things with templates, knew the language well, etc. But then I met OCaml, and now I wonder, "how did anyone ever get by without discriminated unions, closures, currying, first-class functions, GC, a static type systems that actually works, etc.". Now don't get me wrong, several of those features you can emulate in C++, but it is non-trivial to do right, and not very terse, makeing you not want to do it often. And I know Scheme is more powerful, it has a better macro system and continuations, but alas, I may be at that point I mentioned above.
Any language with a top level loop (type, debug, execute code interactively) already has the integration thing down. Languages like LISP and scheme and ocaml already have macro systems.
And by macro systems I don't mean the C like things, I mean the ability to treat code as data and process it like any other data. Wonderful, but rather an old idea.
My my my, what a bad surround system. Might I suggest a 60 point system arranged in a sphere (think bucky-ball or soccer ball). With you idea of 3 channels per point, I am up to 180 channels. And since I use a sphere, I can keep adding points in a semi-normal arrangement.
Bonus points if the room looks like the room in X-men or the astro-something room in STNG.
Oh, and might I suggest just using second order ambisonics? Then you can decode to an arbitrary number of speakers and speaker arrangement, and still only use a handful of channels.
I use to feel this way (though the starwars sets are great). But what lego was realizing was that not all children (or even most?) express their imagination by building. Many express imagination by role playing and such. Thus the more themed and simpler sets.
Recently I was releaved to see mostly basic sets making a comeback; check out the "LEGO Designer Sets" for some good old fashon building sets with mostly basic peices.
> Lots of little things like this, IMHO, make C# better than Java. > Python's better than everything else anyway. *hides*;)
After using OCaml for a while, you start to fill the same way. Type inference, polymorphism that actually works, closures, parameratized modules, anonymous functions, etc all make programming much easier. Other languages have to add hacks (Java: nested classes, generics; C++: templates, anonymous structs that overload () ) to get what ML programmers have for free.
And let us not forget about code speed and memory usage. Go check out the Great Language shoot out (mentioned some time ago here), or better, the Win32 version since that one includes C#. Java implementations are simply embarissingly memory greedy. C# looks much better there, but then look at the OCaml compiler. Code speed rivals gcc.
Yes, but its memory usage is also 40 times higher than python (ok, more like 4x). Take a look at the great computer language shootout for some really scary benchmarks on java memory usage!
Slashdot Mirror
Writing a gtk-to-Qt translator say re-using the LLVM frontend would be a very nice Ph.D. project :)
No it wouldn't. I doubt that would even be a publishable paper in a decent conference. (I appear on both http://llvm.org/developers.cgi and http://llvm.org/pubs/ )
So debian release names are unclassified and Ubuntu release names are classified. Makes sense.
Further proof that no one cares: Steam's Hardware Survey March 2010. Most prevalent resolution amongst gamers? 1280x1024, at 19%. Second place is 1680x1050, at 18%. Neither of those are particularly high, with the highest resolution in the survey being 1920x1200 at 6% and "Other" is only 3.4%.
His point, and I have to agree with it, is that I had 1280x1024 in 1995 and 1600x1200 a few years later. And incidentally the displays were brighter and had a larger color gamet.
Besides when his eyes go in a few years he won't care about the high resolutions anymore.
Because he is staring at low resolution screens! For the sake of your eyes, get a high resolution display!
I read it again and think that mentioning that you are interviewing there is no problem at all. That you are you know before you sign the NDA, and ANYTHING you know before you sign the NDA is explictly NOT covered under the NDA.
What you are not allowed to do in the section mentioned there is use the fact that you had a meeting with google in your marketing or press releases. It does not say you can never mention google, as the summary indicates. For individuals, the most likely and probably only implication of the section is that you might not be able to use an employment offer from google as a bargening chip with other companies, but this isn't clear to me, as I don't know if that can be considered in the classes of things you cannot disclose from other sections (or if by nature a job offer to you does not ential a duty of confidentiality on your part since it isn't related to trade secrets or anything else explicitly mentioned).
I'm as skeptical as the next nerd, but it's still essential for respected scientists to conduct the tests, do the math, and come up with an answer, even if just to debunk it formally. Anything less is a negation of what science is supposed to be about, and reduces scientists to the level of the Holy Office of the Inquisition, condemning a theory without testing it.
No. They want someone to take it seriously, patent the device, hire some people to perform the tests, and try to publish the results. When they submit their paper they will get an instant "jury" to tell them what additional tests they should do and where they were anything less than convincing. It's called peer review. This is just publicity.
It isn't up to the "respected scientists" to refute this, it is up to them (the inventors) to produce enough evidence that their device cannot simply be blown off as a crackpot perpetual motion machine.
Besides, I don't see any "theory" in their published information. So what exactly are the "respected scientists" suppose to test? Anything less than bothering to write up what you did before trying to convince people you are right is worse than a negation of what science is about, it is just lazy and insulting (on the off chance this isn't a crackpot thing).
Everything they ask for would be handled by the normal peer review of a paper submission. It might be a fun paper to review too (some are much more interesting than others, as I am sure you can imagine).
And that is clearly why certain members of the Bush administration (namely Bush) call it nucular instead of nuclear. Opinion polling found this to be a more friendly name.
A while ago I read the RFC. It is very scary. Multihoming as proposed moves things like name resolution into the kernel.
I will grant SCTP does some neet stuff, the best is that it allows independent non-mutually-blocking streams over one connection. It also has state cookies, yum.
SCTP tries to be all things to all people in one protocol. It reads as though they just decided the whole layered protocol thing was overrated and shoved every new feature into this one layer.
Haven't you ever heard of rollback?
On network drivers at least, the driver has such a small interface the driver is put through a full formal methods based proof system to prove the driver doesn't have certain classes of bugs. I hear MS has quite a cluster devoted to model checking network drivers. Granted this doesn't work for all classes, but at least some get fairly rigerous verification. I personally wish Linux drivers got this kind of checking.
No, this will not exist forever (without serious hacking of the physical box). Imagine a world were all interconnects (including to the speakers) are encrypted with a one time key, digital, etc etc, and no licensed box may have analog output. (Obviously, speakers will need to do D to A and have their own amp). Now stop imagining and just wait a few years.
So unless you want to put a mic in front of a speaker (how many op-amps, DAC, and ADC are involved in that path?), you are out of luck with. No line out for you!
Oh my. That some moderator modded the parent "Insightful" rather than 'Funny' says something about how far we are from "News for Nerds". I guess this is now "News for people who think computers can be kind of cool, sometimes".
:) ) was the oldest kernel I ever was doing anything useful with, so I guess I am not a true 'Nerd'.
For those of you still scratching your heads, go do some basic research on how kernels development use to work. Reading LWN.net archives might be a good start.
Ah I remember the old 1.2/1.3 days when 1.2 only supported one instance of my scsi cards, and 1.3.(hundred something) didn't like other hardware. Sorry I think 1.2.x (for some x I don't remember, but could look up
Um, I don't think you know much about Alpha or NT for DEC Alpha. First off, all instructions on Alpha are 32 bits long, and all address 64 bit address spaces. There are no "64-bit instructions" if you mean instruction length and there are no "32-bit instructions" if you mean address mode. Of course when you say "32-bit instructions in parallel" I have no clue what you really mean.
Now, certain compilers and OSes tried to ease the transition by arranging that nothing an application would ever allocate would fall outside the first 32 bits of address space, thus applications that cast pointers to ints (totally broken as they can be different sizes and are on alpha) wouldn't brake due to the developer's incompetence.
If you mean FX32, then we are talking about a binary translator for x86 -> alpha. This allowed you to run x86 win32 binaries on alpha, translating the instructions at run time, and forwarding the library and system calls to the appropriate native alpha OS component (as the OS did not itself run under FX32, but was native alpha code).
Of course, NT for alpha was indeed a total hack. There was special PAL code that made the paging mechanism act more x86 like. PAL code to make exceptions behave the way NT wanted them to. There were plenty of very hackish things about that port, but the whole "n-bit instruction" thing you mention is just gibberish.
I speak this as someone who has run alphas for many years (I have a 21064 still running, amoungst others) and who has recently written an alpha backend to an optimizing compiler.
The first time I read this I saw "gradschool" not "gradeschool" and thought, yup, how true.
Yea, um, sure. Have you ever gone to one of those so called "classes" future MBA's have to take? Thought not. But perhaps you would get bored of learning how to make Excel make graphs. Oh, and think about what that "M" stands for, and realized that most of the torture refered to by grad students (and by this I mean phd students, not masters students) is most apparent after they prelim or qual (depending on the school). Before that, life isn't too much worse off than for a masters student.
"What are the sofware tools of the future going to be?"
I suspect a tool that corrects spelling mistakes will prove to be a useful "sofware" tool.
The cost of buying software is only a small fraction of the cost of using it. The companies you mentioned know this. Perhaps so do the VPs?
I too remember playing with lego. Oh, that was yesterday. Somethings are best kept in short term memory!
And perhaps I am older than the average slashdotter...
>Saying Java isn't cool is like saying Scheme or ML
> isn't cool. It's just a personal preference,
No, Scheme and LISP (as well as modern ML varients) have features that are more powerful than Java. They can express things that cannot be expressed in Java, mainly code generation and modification at runtime, closures (Java has an ugly hack added for AWT), continuations, syntax extention, etc.
It is said that everyone eventually settles on some language on the power spectrum. You look at the features of your language and are reasonably satisfied. You look at simpler/less powerful languages and you wonder "how could anyone program effectively without feature x" (where x is one of you favorite features not available in the simpler language). You look at more powerful languages, and being rather satisfied with your current choice, say "What do I need feature y for? I don't use it now and see no reason too since I can program what I need with my language." (where feature y is a feature the more powerful language has and its supports make a great deal of hoopla over, whereas your language does not).
I was happy with C++, I could do amazingly strange and terse things with templates, knew the language well, etc. But then I met OCaml, and now I wonder, "how did anyone ever get by without discriminated unions, closures, currying, first-class functions, GC, a static type systems that actually works, etc.". Now don't get me wrong, several of those features you can emulate in C++, but it is non-trivial to do right, and not very terse, makeing you not want to do it often. And I know Scheme is more powerful, it has a better macro system and continuations, but alas, I may be at that point I mentioned above.
Any language with a top level loop (type, debug, execute code interactively) already has the integration thing down. Languages like LISP and scheme and ocaml already have macro systems.
And by macro systems I don't mean the C like things, I mean the ability to treat code as data and process it like any other data. Wonderful, but rather an old idea.
My my my, what a bad surround system. Might I suggest a 60 point system arranged in a sphere (think bucky-ball or soccer ball). With you idea of 3 channels per point, I am up to 180 channels. And since I use a sphere, I can keep adding points in a semi-normal arrangement.
Bonus points if the room looks like the room in X-men or the astro-something room in STNG.
Oh, and might I suggest just using second order ambisonics? Then you can decode to an arbitrary number of speakers and speaker arrangement, and still only use a handful of channels.
I use to feel this way (though the starwars sets are great). But what lego was realizing was that not all children (or even most?) express their imagination by building. Many express imagination by role playing and such. Thus the more themed and simpler sets.
Recently I was releaved to see mostly basic sets making a comeback; check out the "LEGO Designer Sets" for some good old fashon building sets with mostly basic peices.
> Lots of little things like this, IMHO, make C# better than Java. ;)
> Python's better than everything else anyway. *hides*
After using OCaml for a while, you start to fill the same way. Type inference, polymorphism that actually works, closures, parameratized modules, anonymous functions, etc all make programming much easier. Other languages have to add hacks (Java: nested classes, generics; C++: templates, anonymous structs that overload () ) to get what ML programmers have for free.
And let us not forget about code speed and memory usage. Go check out the Great Language shoot out (mentioned some time ago here), or better, the Win32 version since that one includes C#. Java implementations are simply embarissingly memory greedy. C# looks much better there, but then look at the OCaml compiler. Code speed rivals gcc.
Of corse, OCaml does better than both. :)
Many of the functional languages I know are also OO. You haven't seen weird until you have a functional object.