The only thing holding me back is gaming. With Vulcan, it looks like things may change. For MS Office, I will run a VM somewhere without network, so even Win 7 after support expires will do.
Scary thing is I've interviewed candidates who claimed to have some RH training as their linux background, yet could not answer some relatively simple command line stuff. I'm assuming/hoping this was some crappy 'Linux training' class rather than any official RH curriculum...
This likely was the official thing. Expect a lot of coders that cannot even compile a "hello.c" in the near future from all those bogus "learn to code" things in addition. Thing is you cannot teach Jonny (or Jane) do code or do system administration with any actual understanding of what they are doing. Those that can will already have started teaching themselves, the others cannot and no amount of "training" can change that.
I agree to that. The thing is, the "Linux Desktop" has been finished mostly before Linux even was around. Except for badly-managed aberrations like Gnome and KDE, most window managers are in maintenance-mode because there is nothing left to do.
Very much so. However there will not be any AI that does deserve the name anytime soon. Sure, we will see more "expert" systems, that can do one very narrow task very well (like playing Go or Chess) and maybe even drive a car competently on good roads in standard conditions, but that is about it. I do agree that the typical programmer is pretty much doomed though either way. They are technicians (think the people that shovel the coal into the boiler) in an age where engineers describe things (and do the remaining implementation themselves because that is faster) and then automation does the production.
Obviously, you have never implemented any non-trivial data-structure under resource constraints and with (soft) real-time, reliability and security requirements. Nothing brain-dead about it at all.
I have done quite a bit of C code for customers the last few years, because they were unable to find anybody else competent. I learned coding in C something like 30 years ago. Sure, things have gotten more complicated, for example, I am also the security architect on that code (and a few other things), but as my skills have grown as well, I do not find this more challenging than 30 years ago.
I agree to that. A good programmer today must be a well-rounded engineer with strong coding skills. But he/she must also be reasonably competent at architecture, design, network technology, system administration, security, writing documents, presenting things, process design, customer communication, etc. This separates those that have a bright future from those that will be unemployed in a few years. Incidentally, this is not the only field where that is happening. I know a graphical artist with a classic university education in the field, and she used to do graphics. Now she also does text, typesetting, printing technology, photography, videos, cost estimation, some project management, etc. And you know what? While the field pays ridiculously low for most working in it, she has no such problems.
My take is that with more and more automation in production technologies, we see a large divide between those that are well-rounded professionals that can learn anything loosely related to their core-skills and then do it reasonably well and those that are basically 1-trick ponies, like the "web application developer" that is not only restricted to web-applications, but also to Java using a specific IDE on a specific platform and cannot even read JavaScript. Those people will first get outsourced, then off-shored, and then replaced by dumb automation (not AI, that one will still be dumber than even the dumbest coder for a long, long time and possibly forever).
The real challenge here is to either keep capitalism working (and I really see no way to do that without some not-so-basic stipend for everybody, people have to be able to buy nice things or capitalism collapses) or replace it with something that can deal well with the fact that long-term something like 80% of the population will not be able work for a living (but will also not need to if the available wealth and production capabilities are shared reasonably). Will be interesting to see how this turns out. I do admit I am very glad that I have some skills that are in high demand and cannot be replaced by anything else though.
Couldn't agree more. Replacing coders in places where they need to understand the problem (maybe 10% of all coding jobs from the abysmal trash most coders produce these days) is not going to happen anytime soon and possibly never. What is going to happen is that the cheap, single-language coders, that have zero understanding of networking, systems, software maintenance, actual software engineering, security, processes, etc. will be recognized as what they are: A factor that drives costs up, often massively so. At that point, good coders will find themselves able to select what work they want to do and what fee they want to ask and the rest will find themselves out of a job. We have far too many coders at the moment, and most of them bad.
We are doing some coding as a side-business when our customers are unable to find coders to implement architectures and designs we do for them. (Can you believe an enterprise with 20k+ IT people has nobody on staff that can code an Apache module? 20k+ Java monkeys, 5 reasonable C++ coders that are restricted to the database world and that is it. Staggering.) We do the coding and any other engineering at full consulting rates and we are planning to raise those rates as more and more customers have coding and IT engineering tasks where they are unable to find anybody competent to do them.
And if you actually believe that bullshit you just wrote, I have some nice shiny new unproven technology that does make grand promises it will never deliver on to sell to you!
Well, now they have 800k people that will not play on their network and are angry at them (and will spread that). How is that better? Apparently, the fatal combination of greed and stupidity (and lawyers as a booster to make it worse) has taken root at Blizzard as well.
"It simply boils down to private servers being illegal" is a lie and a cowardly cop-out. No, private servers are not "illegal". They may be forbidden by Blizzard, but there is no law that says Blizzard cannot allow or tolerate them.
Naaaa, sounds like the first person above script-kiddy level got in. Boring from a technological point-of-view. Even a reasonably done simplistic penetration-test would probably have shown how bad things are. I guess they had no money for that with them all busy getting rich.
With them optimizing profits, they probably had no money for IT security to spare. Save a million, lose a billion (or rather more in this instance). The fatal combination of greed and stupidity at its finest. Will not be the last instance of something this large happening due to non-understanding of IT security.
When the first successful hack costs you everything, learning from experience is not a good strategy. Consulting and listening to some (admittedly expensive, but worth it) real experts may be a good idea.
Last I looked, several hundred dollars would buy a nice randomness generator based on radioactive decay, which you would plug into a USB port. Given that, it's easy to select any desired fraction of people for additional screening. By changing the probability of selection based on how many people had been selected already, it's possible to damp down the random variability at the cost of predictability.
Waste of money. A reverse-breakdown source gives you about half tunneling noise and about half thermal noise, amplified nicely by avalanches. The whole thing costs a few EUR/USD and whatever you want to spend on the USB interface. Add some post-processing and you are good. Incidentally, a Geiger-Mueller tube also has thermal noise, known as the "zero rate", and it also has a time where it cannot detect pulses after a pulse, so you need some post-processing anyways.
First, you are not counting zeros and ones, you are averaging unsigned byte value. That is different, but lets overlook that for now.
Bias is basically any deviation from not having memory and perfect 50:50 distribution. There are a lot of theoretical and concrete tests, but basically they all look at sequences of bits and check some variation or (for the practical tests) approximation of "all sequences of equal length must show up equally often".
Now, as to your example: Calibrating a TRNG to a precision that actually approaches 50:50 and not having any whitening or other post-processing (that in effect does whitening) is in infeasible in practice. The precision needed for your example seems to be roughly somewhere around 0.1%, which is feasible in practice. The deviation between the two runs is large enough for the given sequence length that a systematic source (temperature, e.g.) seems likely.
For a crypto-generator, 0.1% is a large bias. (For an example that this is a large bias in crypto, look at how RC4 was broken.) For a simulation generator, this may also be a rather large bias, depending on use. It causes higher-order correlations between the bits, and that can potentially be very bad. And this single-bit figure is not enough, as bits may be correlated in addition, for example, "10" showing up more often than "01", which will not even show up in single-bit statistics, or an occasional too-frequent short run of "1....1" which is not enough to skew the 0/1 statistics significantly, but which will be a huge problem both for crypto and for simulation.
That runs into a lot of problems, practical, legal, cost, performance and with customer acceptance, otherwise this limited approach would not be done. But sure, in Soviet Russia you certainly search everybody.
Indeed. This thing is part of some sort of campaign (non-organized, is my guess, but still targeted) to make people accept digital fakes of real things as the genuine article. As usual, the motivation is greed. In the same fashion we will eventually get digital "actors", "writers", "game designers", etc. and the quality will be sorely lacking in all of those. A large part of the customer-base will not notice that though.
Fascinatingly stupid and insightless comment. We do actually have a lot of recorded information from that time, and it basically says all the same: Suppression of most science and advancement of society by the church. This makes mots of the recorded information (recorded by the church) tedious and exceptionally boring, but it is there. It is just that nothing much did happen.
Slashdot Mirror
From what I see, this is nowhere near as useful as Cygwin.
The only thing holding me back is gaming. With Vulcan, it looks like things may change. For MS Office, I will run a VM somewhere without network, so even Win 7 after support expires will do.
Scary thing is I've interviewed candidates who claimed to have some RH training as their linux background, yet could not answer some relatively simple command line stuff. I'm assuming/hoping this was some crappy 'Linux training' class rather than any official RH curriculum...
This likely was the official thing. Expect a lot of coders that cannot even compile a "hello.c" in the near future from all those bogus "learn to code" things in addition. Thing is you cannot teach Jonny (or Jane) do code or do system administration with any actual understanding of what they are doing. Those that can will already have started teaching themselves, the others cannot and no amount of "training" can change that.
No idea. I use fvwm, but certainly not on an inflexible and badly structured distro like Red Hat.
I agree to that. The thing is, the "Linux Desktop" has been finished mostly before Linux even was around. Except for badly-managed aberrations like Gnome and KDE, most window managers are in maintenance-mode because there is nothing left to do.
Obvious troll is obvious.
Very much so. However there will not be any AI that does deserve the name anytime soon. Sure, we will see more "expert" systems, that can do one very narrow task very well (like playing Go or Chess) and maybe even drive a car competently on good roads in standard conditions, but that is about it. I do agree that the typical programmer is pretty much doomed though either way. They are technicians (think the people that shovel the coal into the boiler) in an age where engineers describe things (and do the remaining implementation themselves because that is faster) and then automation does the production.
Obviously, you have never implemented any non-trivial data-structure under resource constraints and with (soft) real-time, reliability and security requirements. Nothing brain-dead about it at all.
I have done quite a bit of C code for customers the last few years, because they were unable to find anybody else competent. I learned coding in C something like 30 years ago. Sure, things have gotten more complicated, for example, I am also the security architect on that code (and a few other things), but as my skills have grown as well, I do not find this more challenging than 30 years ago.
I agree to that. A good programmer today must be a well-rounded engineer with strong coding skills. But he/she must also be reasonably competent at architecture, design, network technology, system administration, security, writing documents, presenting things, process design, customer communication, etc. This separates those that have a bright future from those that will be unemployed in a few years. Incidentally, this is not the only field where that is happening. I know a graphical artist with a classic university education in the field, and she used to do graphics. Now she also does text, typesetting, printing technology, photography, videos, cost estimation, some project management, etc. And you know what? While the field pays ridiculously low for most working in it, she has no such problems.
My take is that with more and more automation in production technologies, we see a large divide between those that are well-rounded professionals that can learn anything loosely related to their core-skills and then do it reasonably well and those that are basically 1-trick ponies, like the "web application developer" that is not only restricted to web-applications, but also to Java using a specific IDE on a specific platform and cannot even read JavaScript. Those people will first get outsourced, then off-shored, and then replaced by dumb automation (not AI, that one will still be dumber than even the dumbest coder for a long, long time and possibly forever).
The real challenge here is to either keep capitalism working (and I really see no way to do that without some not-so-basic stipend for everybody, people have to be able to buy nice things or capitalism collapses) or replace it with something that can deal well with the fact that long-term something like 80% of the population will not be able work for a living (but will also not need to if the available wealth and production capabilities are shared reasonably). Will be interesting to see how this turns out. I do admit I am very glad that I have some skills that are in high demand and cannot be replaced by anything else though.
Couldn't agree more. Replacing coders in places where they need to understand the problem (maybe 10% of all coding jobs from the abysmal trash most coders produce these days) is not going to happen anytime soon and possibly never. What is going to happen is that the cheap, single-language coders, that have zero understanding of networking, systems, software maintenance, actual software engineering, security, processes, etc. will be recognized as what they are: A factor that drives costs up, often massively so. At that point, good coders will find themselves able to select what work they want to do and what fee they want to ask and the rest will find themselves out of a job. We have far too many coders at the moment, and most of them bad.
We are doing some coding as a side-business when our customers are unable to find coders to implement architectures and designs we do for them. (Can you believe an enterprise with 20k+ IT people has nobody on staff that can code an Apache module? 20k+ Java monkeys, 5 reasonable C++ coders that are restricted to the database world and that is it. Staggering.) We do the coding and any other engineering at full consulting rates and we are planning to raise those rates as more and more customers have coding and IT engineering tasks where they are unable to find anybody competent to do them.
And if you actually believe that bullshit you just wrote, I have some nice shiny new unproven technology that does make grand promises it will never deliver on to sell to you!
Well, now they have 800k people that will not play on their network and are angry at them (and will spread that). How is that better? Apparently, the fatal combination of greed and stupidity (and lawyers as a booster to make it worse) has taken root at Blizzard as well.
"It simply boils down to private servers being illegal" is a lie and a cowardly cop-out. No, private servers are not "illegal". They may be forbidden by Blizzard, but there is no law that says Blizzard cannot allow or tolerate them.
I am referring to the idea that bad IT security would mean "better transparency".
Naaaa, sounds like the first person above script-kiddy level got in. Boring from a technological point-of-view. Even a reasonably done simplistic penetration-test would probably have shown how bad things are. I guess they had no money for that with them all busy getting rich.
Your comment is so stupid, it is staggering.
With them optimizing profits, they probably had no money for IT security to spare. Save a million, lose a billion (or rather more in this instance). The fatal combination of greed and stupidity at its finest. Will not be the last instance of something this large happening due to non-understanding of IT security.
When the first successful hack costs you everything, learning from experience is not a good strategy. Consulting and listening to some (admittedly expensive, but worth it) real experts may be a good idea.
Last I looked, several hundred dollars would buy a nice randomness generator based on radioactive decay, which you would plug into a USB port. Given that, it's easy to select any desired fraction of people for additional screening. By changing the probability of selection based on how many people had been selected already, it's possible to damp down the random variability at the cost of predictability.
Waste of money. A reverse-breakdown source gives you about half tunneling noise and about half thermal noise, amplified nicely by avalanches. The whole thing costs a few EUR/USD and whatever you want to spend on the USB interface. Add some post-processing and you are good. Incidentally, a Geiger-Mueller tube also has thermal noise, known as the "zero rate", and it also has a time where it cannot detect pulses after a pulse, so you need some post-processing anyways.
First, you are not counting zeros and ones, you are averaging unsigned byte value. That is different, but lets overlook that for now.
Bias is basically any deviation from not having memory and perfect 50:50 distribution. There are a lot of theoretical and concrete tests, but basically they all look at sequences of bits and check some variation or (for the practical tests) approximation of "all sequences of equal length must show up equally often".
Now, as to your example: Calibrating a TRNG to a precision that actually approaches 50:50 and not having any whitening or other post-processing (that in effect does whitening) is in infeasible in practice. The precision needed for your example seems to be roughly somewhere around 0.1%, which is feasible in practice. The deviation between the two runs is large enough for the given sequence length that a systematic source (temperature, e.g.) seems likely.
For a crypto-generator, 0.1% is a large bias. (For an example that this is a large bias in crypto, look at how RC4 was broken.) For a simulation generator, this may also be a rather large bias, depending on use. It causes higher-order correlations between the bits, and that can potentially be very bad. And this single-bit figure is not enough, as bits may be correlated in addition, for example, "10" showing up more often than "01", which will not even show up in single-bit statistics, or an occasional too-frequent short run of "1....1" which is not enough to skew the 0/1 statistics significantly, but which will be a huge problem both for crypto and for simulation.
That runs into a lot of problems, practical, legal, cost, performance and with customer acceptance, otherwise this limited approach would not be done. But sure, in Soviet Russia you certainly search everybody.
Indeed. This thing is part of some sort of campaign (non-organized, is my guess, but still targeted) to make people accept digital fakes of real things as the genuine article. As usual, the motivation is greed. In the same fashion we will eventually get digital "actors", "writers", "game designers", etc. and the quality will be sorely lacking in all of those. A large part of the customer-base will not notice that though.
[clutches pearls]
That sounds painful....
Fascinatingly stupid and insightless comment. We do actually have a lot of recorded information from that time, and it basically says all the same: Suppression of most science and advancement of society by the church. This makes mots of the recorded information (recorded by the church) tedious and exceptionally boring, but it is there. It is just that nothing much did happen.
There is a sucker born every minute....
Seriously, anybody that ever trusted Google is nuts. It is an evil advertising company.