And maybe more importantly, if it disagrees with the humans, can we figure out why? This is a problem with a lot of machine learning applications, but there aren't many where understanding the decision-making process is more vital than it is here.

Agree.

Over the past year I've (for the first time) used Mac OS X on my laptop, I find it much less useful, and frankly much less user friendly, than Gnome 3 (and even Gnome 3 hides too much information because it assumes its users are technophobes).

One can understand Microsoft and Apple designing user interfaces primarily for technophobes, because in the modern world the majority of their users are people for whom the full power of a computer system is too complex for them to understand, much less use; and, seeing that they have in effect a duopoly, the fact that their more technically able users are not well served by their user interfaces doesn't matter, because there aren't enough of us to be a significant market, and most of us will be told what to use at work in any case.

But I really don't understand the Gnome designers' reasons for hiding so much, for making even moderately technical things so awkward. In practice, almost everyone who chooses to use Gnome is a geek. Having said that, if it really annoyed me I could either switch to something else or get under the hood and modify it, and I don't.

For me, Gnome 3 works with niggles. MacOs X is really annoying, but I can use it. Windows 7 is tolerable. Windows 10? Just let's not go there.

Then it's poor responsive design.

Seriously, there is a limit to the width of a column of text that it's comfortable to read, so for continuous text on large screen there may be reasons for having large amounts of whitespace. And, again, for continuous text, having a proportion of white space around the text is easier on the eyes. There can be good ergonomic reasons for using significant whitespace in design.

Good responsive design is hard; to have the same page layout on a two inch wide mobile phone screen as on a 24 inch monitor, and have it attractive and easy to work with on both requires a great deal of thought, and often some compromise. Making the compromises at the small end of the range doesn't work because on a very small screen pages that are not well adapted are completely unusuable, whereas if you make the compromise at the big end of the range you end up with a page that looks ugly but still works.

But the challenge of responsive design is to respond to a wide range of screen sizes and be functional and elegant on all. It's a significant challenge, and too many designers design to one fixed size or a small range of fixed sizes.

He says. On the internet.

If it's true that the passwords have been harvested by malware which uploads the victim's browser's password cache, then this is not just Twitter. It's every site you use. The lesson, if you create websites which require authentication, outsource the authentication function to OpenID providers who have three factor authentication (e.g. Google) - or implement three factor authentication infrastructure yourself, which is not trivial.

It had been sitting on the bookshelf since before Christmas; I picked it up at the weekend and pressed the menu button and... still working.

I don't actually use it much!

... which was about as expensive twenty years ago, when I bought it, as an Apple Watch is now.

Every four years, it needs its battery changed. And that's all.

'Smart' watches are a bit like DAB radio, or, in their day, WAP phones. They're not horrendously expensive, but the user experience is just so much worse than the technology it replaces that no-one's going to buy it. I don't want to take my watch off and recharge it every night. And there is no 'killer app' that I've seen so far that is better on a wrist display than on your phone.

The Pebble, with its low energy monochrome display, is probably a better device than the Apple, but all this generation of 'smart' watches are bulky, ugly, made of non-premium materials, and will have short life; and they're competing on price against beautifully made precision mechanical watches which do the primary job (time keeping) equally well but are built to last a lifetime and require very infrequent attention.

There may be better fourth or fifth generation smart watches in about five years time which compete on quality and charge duration; there may, one day, be a killer app. But at present I see no compelling reason to buy.

It downloads unknown executable code from the Internet, and then executes it. Fortunately the Internet is a very safe place on which no-one would ever dream of posting malicious code.

The nature of mobile devices is that the data connection is of varying quality and reliability, and in some places either slow, intermittent, or non-existent. Relying on apps in the cloud means that your device is fundamentally a brick whenever you go into an underpass or a basement or into mountains and forests. Of course, if you live your whole life in urban areas this may not matter, but for many people it does. Not that this is a criticism of Cortana, or of Microsoft. Google's and Apple's voice services are cloud-backed as well. But for 'the network is the computer' to work, the network has to be ubiquitous and immanent, and for mobile devices it isn't.

What happens to your 'standard' linked lists solution when you have ten values scattered over an array which is 1000! (factorial 1000) in each dimension? For most genuinely sparse arrays, a hashmap is a better approximation of an efficient implementation. Of course, there will be corner cases where you want to do something different, but linked lists strike me as an extremely poor solution except in arrays where more than about 10% of cells have data.

I should have read the linked questions before replying...

Stupid, stupid, STUPID! Why have numRows and numCols in a sparse array? Things with unnecessary, arbitrary bounds annoy me. My implementation of Conway's Game of Life runs on a sparse array precisely because that allows the world to stretch arbitrarily in any direction a glider goes, limited only by the capacity of the bignum library and the total store available to the program.

And this is how we teach computer science?

Sigh.

Sparse array entries, in general, are not necessarily immutable, although they may be so in this case. Most spreadsheets are implemented as sparse arrays, for example. But your point about the benefit of a map is well made.

Sometimes, on Slashdot, it's impossible to tell irony from stupidity.

Close to where I live are large intertidal mudflats. Every other summer some tourist drives a brand new four by four out there and gets stuck. And then, of course, the tide comes in. When the vehicles are recovered two or three tides later, they are insurance write-offs - the electrics, interior, and engine are all beyond repair.

You do not want to immerse something complex and expensive in salt water unless you really, really have to.

One of those occasions where I wish I had mod points but don't. Mod the parent post up!

Seawater is extremely corrosive. Engineering the rocket engine to survive sudden immersion in seawater when very hot would add a great deal to the complexity and cost (and probably weight). And that's before you add the cost of engineering the rest of the vehicle to resist corrosion.