I have never seen a complete and unambiguous requirement for any real piece of software, and I would wager no other software engineer has either.

I wonder why my experience differs from yours - I have found almost all the applications I use (Firefox, PuTTY, Visual Studio, MS Office, TortoiseSVN) work fine with a simple 200% scaling in Windows 7. I agree that Remote Desktop is a pain.

Vista reverts to "fractional scaling", where it simply does a bilinear upscale of the application window, resulting in a blurry, god-awful mess where nothing was rendered natively.

I have to say I've never seen this on either XP or Win7. Perhaps it helps to have exactly 200% scaling so everything can be enlarged exactly. But as far as I can see, programs that aren't scaling aware (such as the command prompt window) are just rendered unscaled. Maybe it is because I have Aero turned off.

I've just gone through interviews at Google and Apple.

At Google, I was asked mainly theoretical questions - big-O, maths/stats, etc. And one "real" architecture/design question at the end. There were 5 interviewers and maybe 7 questions, sometimes 2 per interviewer but usually just 1 that lasted the whole hour. According to my recruiter before the decision, it was maybe 50/50 that I'd get an offer, and I did very well on the real-system design question (by inference, not so well on the others :). I didn't get the job.

At Apple, I had a seven-hour interview with seven interviewers. There were many many questions, far too many to easily remember categories, but they were all focussed on things I might end up doing, or problems that I might end up encountering. I got the job. I guess I do better with "real world" issues than the "consider two sets of numbers, one is ... the other is ...) type.

I have the self-confidence^W^W arrogance to believe I'm an asset to pretty much any company out there, but interview processes are nothing more than a gamble. Sure you can weed out the obvious under-qualified applicants, but frankly (unless the candidate is lying, and in the US that's a real no-no, in the UK padding your CV seems to be sort of expected...) that sort of candidate ought to have been pre-vetoed by the recruiter before getting to the interview.

I've yet to see the interview that guarantees a good candidate will do well. It's all about preparation: can you implement quicksort or mergesort right now, without looking it up ? The algorithm takes about 20 lines of code... Some interviews will require you to have knowledge like that; others are more concerned with how you collaborate with other candidates; still others are concerned with your code quality (I've seen a co-interviewer downmark a candidate for missing a ; at the end of a coding line. I wasn't impressed ... by the co-interviewer. But that's another story); still others are ... you get my point. Whether you do well or not can depend more on the cross-intersectional area of the interviewers style and your own credo than any knowledge you may or may not have.

So go in there expecting to be surprised, prepare what you can, be prepared to do wacky things to please "the man" interviewing you. For a good candidate, over a large number of interviews, you'll do well. The problem is that we often want a specific job, and we get depressed by the first dozen or so failed interviews. There's nothing more you can do than pick yourself up and try again. It's instructive to note that second-interviews at companies often go better than first-interviews, possibly because you're forewarned about the style a bit more, and therefore a bit better prepared...

Not true. Windows can quite easily double resolution - just set your font size to 200% in Control Panel. Everything comes out twice as big but fonts are smoother. This has worked since Windows XP at least. I run it this way on a 3840x2400 monitor, so the effective desktop size is "1900x1200".

Its completely broken with perfectly legitimate local DNS URLs.

Can you give an example of a 'perfectly legitimate URL' that doesn't work? (Hint: all URLs begin with a scheme such as http: or file:)

In my first year at college, I pushed a friend (by accident) through a plate glass window. The college authorities fined me £50 and asked me to be more careful. [friend] was taken to hospital, lost a small slice of an ear IIRC but was otherwise ok.

We were sitting in the college bar, pretty drunk, and there were these thick radiators that ran along the windows which people sat on. [friend] had slid down between the radiator and the plate glass (10' x 10') window, and I thought it'd be a fine idea to get him stuck down there, so pushed him down as hard as I could...

Plate glass windows make a lot of noise when they break...

I do remember grabbing hold of him and pulling him back as soon as it happened, which may be why he still talks to me :) It may also be why he didn't get a sheet of glass through his neck, Exorcist-style.

The dean in charge of my hall-of-residence was particularly scathing when he found out I was studying physics at the time, various comments about the fragility of glass were made, but his (and the college's) attitude was "shit happens around students". The fine was their way of saying "don't be a dick, again".

Of course, this was the UK, not the US. I also wrote a networked virus without ending up in jail...

People repeat this meme that desktop OSes don't support high resolution displays, but it's not really true. At work I have Windows 7 and several 3840x2400 displays; just setting 200% font size works fine in all the applications I use.

Well, yeah, surely 4000x4000 is 4k (like IMAX), and 3840x2160 is only about 2k measured conservatively.

"Riddle me this: what is it that is always coming, but never arrives?"

So you have your nicely written function of type int -> int, but now you want to add some trace statements while it executes, but you can't because it is purely functional. Then someone helpfully suggests monads. Now all you have to do is rewrite it to be of type int -> IO int, and rewrite all of the calling functions to be of monadic type too, and so on all the way up to the top level of the program...

The fundamental problem is that you need some way to separate the program itself (which is purely functional) from whatever scaffolding and diagnostic code you put in to help test and debug it. In imperative languages you can happily mix the two together but in pure functional programming you can't, and the language environment has to provide explicit support for trace messages and the like.

IMHO Windows 7 requires Classic Shell and some messing around in Control Panel to make it usable. Essentially, with each new Windows release there is another half hour of setup you must do to get rid of the latest bright ideas from Microsoft and get back to the basic but usable interface circa 1999.

That US remake of House of Cards isn't too bad.

I remember on the Archimedes using PC emulators, both Acorn's one (which was oriented more towards business applications) and Dave Lawrence's FasterPC (optimized for older DOS games using 320x200 256 colours). Both only emulated an 8086 (or 80186 in FasterPC's case), so they were good for Civilization or SimCity but not newer things. I mention them because, to squeeze decent performance out of an ARM running at speeds as slow as 8MHz, they were hand-written in optimized assembler. This was the older 'ARM26' instruction set where the program counter and flags were contained in a single 32-bit register, so the code would need modification to run on recent ARM processors; and in any case the I/O part of the code will be totally different. But that doesn't stop me wondering about the core CPU emulation and how fast it might go on modern hardware.

So Windows 8 doesn't have the Aero Snap, Aero Peek and other animated window management thingies? I can understand that Microsoft marketing stopped calling it "Aero", and that Aero Glass (the odd half-transparent effect on window borders) has been dropped, but I thought the thing itself (whatever it's now called) had been kept.