The ability to block internet access would effectively block ads. On the plus side, there are plenty of firewall apps in the Play store (though they do require you to have a rooted device).

I just googled it, and it absolutely does appear in that special box. (Interestingly, this is despite it being the #2 result, with #1 being the PBS website.)

I disagree. Take a look at D templates (and compile-time function evaluation) - that syntax will be easily readable to anyone familiar with a C family language, because it's the same syntax (just with 'static' or 'pure' peppered everywhere).

C++ template meta-programming is hard for two reasons. The first is that templates use a syntax of their own, which is one more thing to learn. The second is that the template language is functional, which is quite alien to most C++ programmers. D uses native D syntax to do the same thing (with the requirement that functions be pure), and that's much easier to read.

That's a bit of a strawman. IMO, the real issue is that people see it as a false dichotomy between a good text editor (Vim, Emacs, etc.) and a kitchen sink of tools attached to something little better than Notepad with syntax highlighting + autocompletion (e.g. Visual Studio). In that comparison, the text editors win because the vast majority of programming comes down to text editing (and possibly thinking, but you don't need a computer for that), and typical IDEs are pretty mediocre at that.

Note that it's a false dichotomy - you can get the benefits of both, either by integrating a proper text editor into your IDE (e.g. Eclim adds Vim to Eclipse) or by adding many plugins to your text editor - I use Vim with 53 plugins, and I have every feature on your list (although hovering over a variable is replaced with typing a shortcut while the cursor is over it). You only need about 10 to get most of the benefits though, especially if you only work with one or two languages.

I've got a pretty good idea why this is happening. The old API for tray icons didn't handle the case where you had multiple systrays correctly (e.g. dual monitors) - the icon would only appear on one of them. The new API fixes this, but as you have noticed not everything supports it. The best solution is probably for someone to go around submitting patches to all the projects using the old API, but that's obviously going to take some time...

I've found avahi works quite well for local/P2P name resolution (at least, much more reliably than SAMBA + winbind). Now admittedly I started using it long after it had stabilized, but I would like to know what you were using instead.

Because parallelism is good for performance, or so the argument goes.

The real issue is that there's no way to reproduce that partial-ordering. IMO, there should be some kind of debug facility to restrict it to sequential startup, then try all possible orders. (Yes, factorial complexity is never fun, but at least it would provide a means to debug this kind of problem. Making open source software easy to debug (or at least capture logs for) is hugely important, because that's where most of your drive-by patches come from.)

It exists, but isn't particularly well known.

IMO, feminism and masculinism are both inherently flawed, in that they only focus on one set of symptoms and don't try to fix the underlying issues. Both genders have issues, but the moment you try to draw attention to this people tend to take it as a competition / zero sum game. The real solution is something like egalitarianism, where the focus is on not discriminating against people regardless of their gender, etc.

Speaking as someone who only started using KDE after 4 had stabilized, what did they break in Kmail?

If your context switches are too slow, the correct solution is to fix the kernel or add syscalls to reduce the overhead (see sibling post). Moving parts of your application into kernel-space is bad design no matter how you look at it. (Besides, wasn't it only a few years ago they had a vulnerability in their kernel-mode font driver?)

IIRC, there is one difference: how they respond to read errors. Consumer drives will keep trying to maximise the likelihood of a successful read, while the enterprise ones will just fail immediately since they're expected to be RAID, so there's another copy of the data and taking longer to reply just kills the throughput.

Serious question: how does one meaningfully distinguish between hardware and software patents? As I understand it, software is supposed to be unpatentable because it is just math, but the same could also be said of Widlar's negative feedback amplifier, since the mathemathical models by which transistors function were well established at that point.

The best argument I've heard against software patents is that they inhibit interoperability (e.g. the MPEG patents), but that's not specific to software - the same is true of Apple's magsafe connector.

To be fair, Arch is a distro for people who are fine with things breaking all the time, which is what you'd expect of a bleeding edge rolling release distro. A review of someone who spent a year running Ubuntu on their laptop would be far more realistic in terms of what a casual user would experience.

If they offer average pay, they'll get average employees. If they offer above average pay, they miss out on exceptional employees. And if they offer exceptional pay, they'll likely go bankrupt as most of their employees will not be exceptional.

Negotiation exists because there's no objective way to evaluate the value of the employee to a company before they've been hired. If someone can get twice the work done (and can demonstrate this), they can justifiably demand twice the pay. Of course, the subjectivity is a double-edged sword, because it means that individual prejudices can affect the hiring process.

One way to solve this problem is to handle all negotiation through a well-defined algorithm. The would-be employee shouldn't even interact with a person for this part of the process, just with a webpage. Strong AI is obviously impractical, but you could probably do a pretty good job of predicting performance if you managed to trawl a big enough dataset for some key statistics.

Another approach is what this poster suggested, where pay (above a base salary) is determined by one's peers. In that case, the individual prejudices are averaged (which ideally negates them), and the valuation of the employee is done by their peers. You'd have to be very careful about how you implemented it though, as you run the risk of creating some major social/political problems with that approach.

You could avoid that by setting master.rebase for that repo, or setting branch.autosetuprebase=always globally. But even if you did accidentally merge instead of rebasing, the merge commit will contain the hashes of both parents, so it's simple enough to reset the branch.