You're right, but it's not always the devices within the same product category. A lot of stuff that's in consumer devices begins life in very niche applications (e.g. military or medical devices) to get the first bit of R&D funding and then needs another big chunk to become cheap enough for consumer devices.

You misunderstand. It might not be free for you, but it's free for the fish...

It's not clear that Apple could survive in isolation. A lot of their components are only as cheap as they are because of other lower-margin companies paying a big chunk of the R&D costs. When Apple was using PowerPC processors and were the only customer for IBM or Motorola for a particular chip, they found it very difficult to compete. They're designing their own ARM cores now, but they're benefitting enormously from the thriving ARM software ecosystem.

The problem is that the costs outweigh the revenue. Reducing costs in a way that also reduces revenue is probably not a good solution.

The overall compile-link step is roughly the same (although LTO changes it a bit), but the compilation process has changed hugely in the last 20 years. Dealing with code 'hand optimised' by people who still have a mental model of how PCC compiles code is a constant source of pain.

Why would they do that? If you're a cosmetics company and you can buy a startup that owns the patents on a technique that actually works, then you'd be stupid to keep competing on a level playing field when you could be the only company that's selling the real thing. Even if you multiply your normal profit margin by a factor of ten, you're still going to be selling huge quantities and raking in the money.

The problem with these conspiracy theories is that they assume that people with large entrenched interests and lots of money somehow have an aversion to turning their big pile of money into an enormous pile of money.

Then you're probably not going into an establishment where the only drinkable beverages are laced with sugar and milk to disguise how bad the coffee is.

BT is required to allow third parties to install equipment in the exchanges ('local loop unbundling'), and while most of the companies that take advantage of this are small local affairs, TalkTalk has quite a lot of coverage on LLU exchanges. Since BT won't sell naked ADSL lines, they've priced themselves completely out of the market in areas with Virgin Media coverage.

Well, sort of. Clause 7 could be interpreted as a patent license, in that if you knowingly distribute code that violates your patents then you are violating the license if you don't also include a patent grant. In v3 it's more explicit precisely because it was ambiguous in v2. It's up to the court to decide whether this ambiguous license is a license.

In the book, Pris and Rachel are the same model and identical. Rachel seduces Deckard so that he will have an emotional reaction to Pris, which will slow him down. In the film, she seduces him because the film wanted some implication of sex at that point.

In the book, there's a lot of ambiguity about whether Decard is human, but only in his mind. There's also another bounty hunter who fails the empathy test when asked about androids, but passes it when asked about humans. The 'electric sheep' line from the title is a reference to the fact that, after almost destroying themselves in a nuclear war, humans are expected to keep pets to demonstrate their empathy and it's been engrained in the survivors as the most important character trait, yet the manipulate their emotions with mood organs and via shared religious experiences. Most of the subtext was lost in the movie (fortunately, so was the weird subplot about the police precinct populated entirely by androids, who believed that they were humans hunting androids).

The problem is that they don't differentiate categories well. Having bought some kind of computer thing means that I might be interested in buying some kind of computer thing again, but having bought one hard drive probably doesn't mean that I want another very similar (but not identical) one soon. In books, it's very different - if I've bought one novel then I probably want to buy another very similar (but not identical) one next time I shop. The same is true for a lot of things on Amazon - DVDs, CDs, and even clothes - and so the algorithm works pretty well overall, it just fails laughably in some cases (ah, you've bought a USB flash drive, do you want to buy a different USB flash drive with the same capacity?).

Even without direct enumeration, it's still relatively easy to find. The object / embed tags can be nested for fallback and the resource is only requested if you have that plugin installed. You can provide a load of 1px objects with different nesting and just check in the server which cookies show up in the requests.

It's easy to prevent. The browser should only expose a whitelisted set of system fonts to the web, which would then tell you nothing that wasn't in the user agent string to start with. With the widespread support for Web Open Font Format, it's easy for designers to provide additional fonts if they want to use them. I don't want something random on the web to be rendered in, for example, the Quake font, just because I happen to have it installed - it's almost certainly not what the developers intended, and if it is then they should use a .woff file.