You need a base of unskilled workers to finance the society, or otherwise expect to be robbed in taxes and negative interest for the gov to hand them subsidies.
So your argument is that we're going to be paying unskilled people anyway, we may as well make them do something? Personally, if we're going to pay them anyway then I'd rather that we build efficient systems and treat occupying the time of people who can't contribute usefully to society as a separate problem.
Here's the thing: DRM doesn't stop any of that. If you want to create a YouTube channel full of Hollywood films, you can find thousands of them on usenet, on any torrent web site, and so on. There are many other things (notably, copyright laws) stopping people from this. DRM only inconveniences people who have paid for the content, everyone else gets the DRM-free bersion. If anything, removing the DRM could make it easier to catch people who do this, because Netflix could add a steganographic watermark to each stream and make it possible to identify whoever had originally downloaded it.
From my personal experience, all of the books that I've written are available DRM free, because my publisher (Pearson) found that DRM reduced sales too much to be worthwhile. The PDF of my first book was sent by someone to a mailing list that a lot of the potential target market subscribed to. I was, understandably, a little bit cranky about it, until a few months later when I realised that it had, if anything, had a small positive effect on sales.
Indentation conveys information about the scoping depth, alignment is aesthetic. All lines in a block will have the same indentation, but wrapped lines may have some additional spaces at the start to align them with something from the same expression (e.g. aligning all arguments in a function call).
I have my editor set up to display tabs, so I get a nice coloured shape for the indent level on the left. Spaces to the left aren't coloured, so I get a stronger visual clue for indentation than alignment. You can't do this if you're using the same character for indentation and alignment.
Tab is a character that means 'indent one level to the next tabulator'. It doesn't mean 'advance n spaces'. The width is not encoded by the character, so devs can freely select anything from one to eight spaces, depending on their preferences.
So you are seriously saying Netflix could operate without DRM?
Yes. Netflix DRM does absolutely nothing to decrease piracy. If Netflix provided plain.mp4 downloads (perhaps rate limited to prevent people from trying to download their entire catalogue) then their service would still work. I'd actually subscribe to them (I don't now), because I'd be able to watch their content on the FreeBSD media centre box connected to my projector.
Why doesn't someone operate a service like Netflix with no DRM then?
Because the studios won't license their content to Netflix without DRM (they also wouldn't let iPlayer stream films without at least a token attempt at DRM, even though it was trivial to break). They have failed to learn the lesson of the music industry and are still buying the argument that it decreases piracy as a cover for allowing companies like Netflix to control their channel. Netflix licenses their DRM to a load of set-top-box makers and so on, meaning that there are a huge number of devices that can watch Netflix content. That's a big barrier to entry for a new startup to overcome. If the studios would license their content for DRM-free download, you'd see a load of Netflix competitors spring into being.
if someone is essentially offering some sort of rental model, maybe with access to a large library of content for a specific period at a price far lower than buying a permanent copy of each of the relevant works accessed during that period, then there needs to be some system so customers can't just blatantly take advantage.
So limit downloads. I'd happily pay a subscription fee that limited me to, say, 30 hours a week of video downloads. Even with an unlimited download ability for DRM'd content, I'd still have an incentive to keep up the subscription, because a lot of the value I get is from access to new content. If I sign up and download everything in their catalog then cancel my subscription, I'll have enough to watch for the rest of my life, but I won't be able to watch films and TV shows produced the following year, and that's what I'm most willing to pay for - older TV shows are so cheap on DVD now that it's better to buy the boxed set than stream them.
By including DRM in the standard, you allow everyone to implement the exact same thing, and make it universally available on all devices.
Except that's not possible. DRM relies on secrets, not simply for keys, but in the implementation. You cannot have an open source implementation of DRM, because anyone can simply modify the code to remove the encryption and make the unencrypted stream available. You can standardise a mechanism for plugging proprietary DRM modules into the web, but you're still reliant on the vendors to provide them for your platform of choice. If you're a minority platform, then you're still screwed.
DRM on music was killed by Apple's control over the DRM on the iPod and the iTunes Music Store. They were able to lock all competitors out of providing DRM'd music that played on the iPod. No one wanted to buy DRM'd music that they could only play on a Windows computer or a Zune, so Apple has a very strong bargaining position with the studios. Their only way around this was to allow other stores to sell DRM-free downloads.
It's important to remember that there are three participants in media distribution:
The studio that produces the content.
The consumer that buys the content.
The distributor, who sells the content.
We often conflate the studio and distributor, but DRM is something that benefits distributor over the studio: it allows them to lock customers in and strengthens their negotiating position, because whoever controls the DRM platform is the gatekeeper for the consumer's device. The best way to end DRM is to allow a single company, such as Netflix, to gain so much control over the channel that the only way for the studios to break this control is to allow their competitors to sell without DRM.
Hi, welcome to 2017. AI is now defined by the media to mean 'thing using algorithms'. In related news, algorithm is now defined to mean 'scary thing the reader probably doesn't understand'.
Asymptotic complexity comes up a lot if your code is taking untrusted input, because if an attacker can make your average-case-linear algorithm exhibit its worst case n^2 behaviour then you've created a DoS vulnerability.
It probably won't be done in hardware on a modern GPU. It's been over a decade since that kind of thing has been accelerated with custom hardware. It might be implemented as a pixel shader program that the driver will run, but it's often faster to do it on the CPU than the GPU because you do a lot more than a single fill when drawing a 2D scene and the overhead of copying the data back for the step that you're not offloading to the GPU is more than the overhead of doing everything on the CPU and shipping the final result to the GPU. On-die GPUs are starting to change this, however, and so my answer will also be wrong in a couple of years.
If you answered that, I'd then ask you on what kinds of data set it would be appropriate to use your favourite language's built-in sort and on which it wouldn't as the follow up, and I'd ask you to justify your answer. For example, on a three-element array in a critical path for performance, calling a built-in sort function is almost always wrong (though not necessarily in C++, where the standard library version may be inlined). Other non-toy languages usually document the asymptotic complexity of their built-in algorithms and so even if you don't know what the exact algorithm is, you can find out. If you know that everything in the array is an integer between 1 and 10, for example, then you can perform the sort in linear time, whereas the built-in sort is probably n log(n).
One of the most important qualities in a programmer is understanding that there is almost never a good one-size-fits-all solution, though in many cases there's a good-enough off-the-shelf solution.
Yes. The system should be designed so that everything is recorded and encrypted with one of a rolling set of asymmetric keys. Nothing can be played back without accessing the decryption key, which should not be available to the individual police precinct without oversight. Everything recorded should be stored and, unless needed, deleted a month later. Any decryption keys that are not used within this period should be deleted without ever being released, so even if someone takes an unauthorised copy of the video, they can't decrypt it.
It absolutely should not be up to the judgement of an individual in a highly emotionally charged situation to decide what should be recorded.
It's military and government infrastructure and it turns out that a lot of government infrastructure runs on commodity hardware and software (actually, so does a lot of military stuff). This means that they end up auditing a lot of proprietary and open source code that's widely deployed. This is why Heartbleed was so embarrassing for them: OpenSSL was explicitly on the list of 'critical infrastructure' software that they're meant to be securing.
At least on iOS it's just a usability regression. You can go to settings and remove the permission to access location data at all and as long as you toggle it back before you book a ride. I think recent Android versions are the same, but on older ones it's a choice of privacy violations or uninstalling the app.
You have to wonder a bit whether a boycott would actually help. Uber is currently cheap because they're using VC money to subsidise every ride and making a loss to build up market share. Is it better to use them and cost the company money, or not use them and help ensure that competitors stay in business?
One of the reasons that we teach sorting and searching algorithms is that a huge number of complex domain-specific algorithms include a step that is either searching, pre-sorting, or both. If you don't understand these steps well then you won't have the tools to do a good job at the more complex parts of algorithm design.
Most of the time I've seen this in an interview, and when I've been interviewing, any of that list would be fine. Part of the purpose of whiteboard coding is to see whether the candidate gets hung up on syntax minutiae or whether they focus on algorithms.
It's not that clear cut. Take the SGI example. You sell graphical workstations for $10+K, with a $2-3K mark up on each one. You launch a range of graphics accelerator boards for $500-1000. All of your customers switch to buying commodity workstations from someone else with one of your boards inside. You're now probably not making enough on unit sales to cover your R&D costs and the company goes under. The gamble is that not only will your customers buy these boards, an order of magnitude more people will buy them and you can then use this money to find the $100 version that you can sell to another order of magnitude of customers.
The real danger for SGI was that someone else would start with the $100 board (which was nowhere near competitive with the performance of an SGI machine) and sell it with low margins to enough people (who would never think of buying an SGI machine) that they could then afford to develop the $500-1000 board that would kill the graphical workstation market.
The NSA is evil. Its employees are not doing important work: they are evil.
The NSA is a huge organisation with a dual mission. They are tasked with protecting critical infrastructure and being able to attack other people's critical infrastructure. These are fundamentally contradictory (you find a zero-day vulnerability in something like OpenSSL: mission 1 requires that you disclose it and get it fixed ASAP to protect your infrastructure, mission 2 requires that you keep is secret so that you can use it to attack everyone else) and as such they're largely isolated into different parts of the organisation. Even if you completely disapprove of mission 2, arguing that mission 1 is not important work and is inherently evil makes you seem like an idiot.
When the FBI used a court order to compel Apple to decrypt the San Bernadino shooter's iPhone, Trump encouraged people to boycott Apple if they wouldn't decrypt the phones. This was when Apple was leading a charge to resist court-ordered decryption.
Apple did not resist court-ordered decryption. They cooperated with the FBI even without a warrant to decrypt a single phone, they only objected when the FBI failed to follow their instructions correctly, locked the phone even more, and demanded the ability to decrypt any iPhone without oversight.
I fail to see how nsa has EVER 'kept us safe'. and due to their being untouchable and above the law, we'll NEVER KNOW, either.
Then NSA is a dual-mission agency. They do SIGINT, but they are also tasked with ensuring the security of critical infrastructure. Some of this involves designing crypto protocols, some auditing nominally secure systems, and so on (for example, writing SELinux). It's entirely possible to be working entirely on this kind of thing. You're probably making the world a better place, even if the net contribution of your entire organisation might not be.
Slashdot Mirror
You need a base of unskilled workers to finance the society, or otherwise expect to be robbed in taxes and negative interest for the gov to hand them subsidies.
So your argument is that we're going to be paying unskilled people anyway, we may as well make them do something? Personally, if we're going to pay them anyway then I'd rather that we build efficient systems and treat occupying the time of people who can't contribute usefully to society as a separate problem.
Here's the thing: DRM doesn't stop any of that. If you want to create a YouTube channel full of Hollywood films, you can find thousands of them on usenet, on any torrent web site, and so on. There are many other things (notably, copyright laws) stopping people from this. DRM only inconveniences people who have paid for the content, everyone else gets the DRM-free bersion. If anything, removing the DRM could make it easier to catch people who do this, because Netflix could add a steganographic watermark to each stream and make it possible to identify whoever had originally downloaded it.
From my personal experience, all of the books that I've written are available DRM free, because my publisher (Pearson) found that DRM reduced sales too much to be worthwhile. The PDF of my first book was sent by someone to a mailing list that a lot of the potential target market subscribed to. I was, understandably, a little bit cranky about it, until a few months later when I realised that it had, if anything, had a small positive effect on sales.
Indentation conveys information about the scoping depth, alignment is aesthetic. All lines in a block will have the same indentation, but wrapped lines may have some additional spaces at the start to align them with something from the same expression (e.g. aligning all arguments in a function call).
I have my editor set up to display tabs, so I get a nice coloured shape for the indent level on the left. Spaces to the left aren't coloured, so I get a stronger visual clue for indentation than alignment. You can't do this if you're using the same character for indentation and alignment.
Tab is a character that means 'indent one level to the next tabulator'. It doesn't mean 'advance n spaces'. The width is not encoded by the character, so devs can freely select anything from one to eight spaces, depending on their preferences.
So you are seriously saying Netflix could operate without DRM?
Yes. Netflix DRM does absolutely nothing to decrease piracy. If Netflix provided plain .mp4 downloads (perhaps rate limited to prevent people from trying to download their entire catalogue) then their service would still work. I'd actually subscribe to them (I don't now), because I'd be able to watch their content on the FreeBSD media centre box connected to my projector.
Why doesn't someone operate a service like Netflix with no DRM then?
Because the studios won't license their content to Netflix without DRM (they also wouldn't let iPlayer stream films without at least a token attempt at DRM, even though it was trivial to break). They have failed to learn the lesson of the music industry and are still buying the argument that it decreases piracy as a cover for allowing companies like Netflix to control their channel. Netflix licenses their DRM to a load of set-top-box makers and so on, meaning that there are a huge number of devices that can watch Netflix content. That's a big barrier to entry for a new startup to overcome. If the studios would license their content for DRM-free download, you'd see a load of Netflix competitors spring into being.
if someone is essentially offering some sort of rental model, maybe with access to a large library of content for a specific period at a price far lower than buying a permanent copy of each of the relevant works accessed during that period, then there needs to be some system so customers can't just blatantly take advantage.
So limit downloads. I'd happily pay a subscription fee that limited me to, say, 30 hours a week of video downloads. Even with an unlimited download ability for DRM'd content, I'd still have an incentive to keep up the subscription, because a lot of the value I get is from access to new content. If I sign up and download everything in their catalog then cancel my subscription, I'll have enough to watch for the rest of my life, but I won't be able to watch films and TV shows produced the following year, and that's what I'm most willing to pay for - older TV shows are so cheap on DVD now that it's better to buy the boxed set than stream them.
By including DRM in the standard, you allow everyone to implement the exact same thing, and make it universally available on all devices.
Except that's not possible. DRM relies on secrets, not simply for keys, but in the implementation. You cannot have an open source implementation of DRM, because anyone can simply modify the code to remove the encryption and make the unencrypted stream available. You can standardise a mechanism for plugging proprietary DRM modules into the web, but you're still reliant on the vendors to provide them for your platform of choice. If you're a minority platform, then you're still screwed.
DRM on music was killed by Apple's control over the DRM on the iPod and the iTunes Music Store. They were able to lock all competitors out of providing DRM'd music that played on the iPod. No one wanted to buy DRM'd music that they could only play on a Windows computer or a Zune, so Apple has a very strong bargaining position with the studios. Their only way around this was to allow other stores to sell DRM-free downloads.
It's important to remember that there are three participants in media distribution:
We often conflate the studio and distributor, but DRM is something that benefits distributor over the studio: it allows them to lock customers in and strengthens their negotiating position, because whoever controls the DRM platform is the gatekeeper for the consumer's device. The best way to end DRM is to allow a single company, such as Netflix, to gain so much control over the channel that the only way for the studios to break this control is to allow their competitors to sell without DRM.
What did he do wrong?
Seriously, did you not read the headline? He uses AOL. He clearly can't be trusted with important decisions.
Hi, welcome to 2017. AI is now defined by the media to mean 'thing using algorithms'. In related news, algorithm is now defined to mean 'scary thing the reader probably doesn't understand'.
Asymptotic complexity comes up a lot if your code is taking untrusted input, because if an attacker can make your average-case-linear algorithm exhibit its worst case n^2 behaviour then you've created a DoS vulnerability.
It probably won't be done in hardware on a modern GPU. It's been over a decade since that kind of thing has been accelerated with custom hardware. It might be implemented as a pixel shader program that the driver will run, but it's often faster to do it on the CPU than the GPU because you do a lot more than a single fill when drawing a 2D scene and the overhead of copying the data back for the step that you're not offloading to the GPU is more than the overhead of doing everything on the CPU and shipping the final result to the GPU. On-die GPUs are starting to change this, however, and so my answer will also be wrong in a couple of years.
If you answered that, I'd then ask you on what kinds of data set it would be appropriate to use your favourite language's built-in sort and on which it wouldn't as the follow up, and I'd ask you to justify your answer. For example, on a three-element array in a critical path for performance, calling a built-in sort function is almost always wrong (though not necessarily in C++, where the standard library version may be inlined). Other non-toy languages usually document the asymptotic complexity of their built-in algorithms and so even if you don't know what the exact algorithm is, you can find out. If you know that everything in the array is an integer between 1 and 10, for example, then you can perform the sort in linear time, whereas the built-in sort is probably n log(n).
One of the most important qualities in a programmer is understanding that there is almost never a good one-size-fits-all solution, though in many cases there's a good-enough off-the-shelf solution.
Yes. The system should be designed so that everything is recorded and encrypted with one of a rolling set of asymmetric keys. Nothing can be played back without accessing the decryption key, which should not be available to the individual police precinct without oversight. Everything recorded should be stored and, unless needed, deleted a month later. Any decryption keys that are not used within this period should be deleted without ever being released, so even if someone takes an unauthorised copy of the video, they can't decrypt it.
It absolutely should not be up to the judgement of an individual in a highly emotionally charged situation to decide what should be recorded.
It's military and government infrastructure and it turns out that a lot of government infrastructure runs on commodity hardware and software (actually, so does a lot of military stuff). This means that they end up auditing a lot of proprietary and open source code that's widely deployed. This is why Heartbleed was so embarrassing for them: OpenSSL was explicitly on the list of 'critical infrastructure' software that they're meant to be securing.
That's really only true for entry-level code monkey jobs. Anything else will involve a lot of data structure and algorithm design.
And you consider designing an algorithm to be implementation not architecture? I hope I never use anything that you've written.
At least on iOS it's just a usability regression. You can go to settings and remove the permission to access location data at all and as long as you toggle it back before you book a ride. I think recent Android versions are the same, but on older ones it's a choice of privacy violations or uninstalling the app.
You have to wonder a bit whether a boycott would actually help. Uber is currently cheap because they're using VC money to subsidise every ride and making a loss to build up market share. Is it better to use them and cost the company money, or not use them and help ensure that competitors stay in business?
One of the reasons that we teach sorting and searching algorithms is that a huge number of complex domain-specific algorithms include a step that is either searching, pre-sorting, or both. If you don't understand these steps well then you won't have the tools to do a good job at the more complex parts of algorithm design.
Most of the time I've seen this in an interview, and when I've been interviewing, any of that list would be fine. Part of the purpose of whiteboard coding is to see whether the candidate gets hung up on syntax minutiae or whether they focus on algorithms.
It's not that clear cut. Take the SGI example. You sell graphical workstations for $10+K, with a $2-3K mark up on each one. You launch a range of graphics accelerator boards for $500-1000. All of your customers switch to buying commodity workstations from someone else with one of your boards inside. You're now probably not making enough on unit sales to cover your R&D costs and the company goes under. The gamble is that not only will your customers buy these boards, an order of magnitude more people will buy them and you can then use this money to find the $100 version that you can sell to another order of magnitude of customers.
The real danger for SGI was that someone else would start with the $100 board (which was nowhere near competitive with the performance of an SGI machine) and sell it with low margins to enough people (who would never think of buying an SGI machine) that they could then afford to develop the $500-1000 board that would kill the graphical workstation market.
The NSA is evil. Its employees are not doing important work: they are evil.
The NSA is a huge organisation with a dual mission. They are tasked with protecting critical infrastructure and being able to attack other people's critical infrastructure. These are fundamentally contradictory (you find a zero-day vulnerability in something like OpenSSL: mission 1 requires that you disclose it and get it fixed ASAP to protect your infrastructure, mission 2 requires that you keep is secret so that you can use it to attack everyone else) and as such they're largely isolated into different parts of the organisation. Even if you completely disapprove of mission 2, arguing that mission 1 is not important work and is inherently evil makes you seem like an idiot.
When the FBI used a court order to compel Apple to decrypt the San Bernadino shooter's iPhone, Trump encouraged people to boycott Apple if they wouldn't decrypt the phones. This was when Apple was leading a charge to resist court-ordered decryption.
Apple did not resist court-ordered decryption. They cooperated with the FBI even without a warrant to decrypt a single phone, they only objected when the FBI failed to follow their instructions correctly, locked the phone even more, and demanded the ability to decrypt any iPhone without oversight.
I fail to see how nsa has EVER 'kept us safe'. and due to their being untouchable and above the law, we'll NEVER KNOW, either.
Then NSA is a dual-mission agency. They do SIGINT, but they are also tasked with ensuring the security of critical infrastructure. Some of this involves designing crypto protocols, some auditing nominally secure systems, and so on (for example, writing SELinux). It's entirely possible to be working entirely on this kind of thing. You're probably making the world a better place, even if the net contribution of your entire organisation might not be.