True, and clearly a win.

ROFLMAO. IME, the only thing more painful than maintaining a Windows system over the long term is maintaining a *nix system over the long term.

Let's consider Linux. First, you probably get to choose between a stable or a not stable version of your distro. Choose stable and you're OK as long as you don't need to run any software released in the last 3 years and you're OK with being forced to upgrade the whole OS after maybe 2 years anyway (which will quite possibly trash your entire machine to the point of not being able to boot, or at least breaking minor features like RAID arrays, assuming you actually managed to configure one of those properly in the first place after your distro's "user friendly" installer messed it up completely). Alternatively, choose unstable if you want to run more recent software but don't mind stuff breaking all the time instead of every couple of years on a schedule.

Either way, if you want anything that hasn't got into your distribution's package management system yet, you're almost invariably forced into compiling your own software and manually installing it with makefiles. Those might, if you're really lucky, also offer a make uninstall option that actually does cleanly uninstall. That might, if you're even luckier, still work six months later, as long as no-one inadvertently installed a new version of the manually compiled code over the top to "upgrade" it, or just ran make distclean without thinking leaving you with no idea what make uninstall should have done. In any case, Linux is going to enforce absolutely no system hygiene at any point in this process.

OS X is of course doing much better with a similar foundation, as anyone who has spoken the words "Apple" and "shellshock" in the same sentence over the past few days can testify. Or at least, they'll be able to testify, just as soon as they've finished wiping and reinstalling their botnetted systems, because the patch everyone else had within hours only arrived for Apple gear several days later and long after exploits were widely found in the wild.

You're absolutely right that we should be able to install many programs and uninstall them with no lingering effects. But the idea that the registry is the only thing preventing that on Windows or that *nix systems do better is crazy. The only reason *nix systems don't break more often is that the only people running them are geeks and professionals, and those kinds of people are less likely to install random junk and more willing to dive in and fix internals when stuff goes wrong.

Except there wasn't. Well, there was. A bit. Sometimes. Naturally, this half-baked approach actually made the problems worse.

Even today and with native Windows applications, many aren't very well behaved in following the "standards" here, because Microsoft did such a terrible job of promoting good practices.

Anything that isn't a native Windows application -- including almost every darling of the open source world, for a start -- probably ignores not only the application data directory but also the program files directories and insists on spewing its crap all over your filesystem and environment. Oh, and $DEITY help you if you need to do anything with Cygwin, and $CHORUS_OF_DEITIES help you if you have more than one ported application that requires Cygwin.

It is telling that you can't even schedule a backup of the "official" place to store documents without considerable effort, because Windows itself sets up so many links that most backup tools can't handle them.

And that's before you get idiots like the Chrome team at Google who think it's clever to install executable software in your data directory in order to deliberately circumvent Windows' normal security model, just so their auto-updater can do things it shouldn't without anything silly like troubling the user for permission. I'm always a little surprised that Microsoft hasn't, with considerable and legitimate justification, flagged Chrome's installer/updater software as malware and automatically removed it at some point.

On the bright side, if Microsoft can actually manage to produce an operating system with a sensible filesystem structure and application installation/update/uninstallation tools that actually enforce that structure, they might yet salvage the Windows brand and convince significant parts of their potential market to upgrade again.

You can make Windows systems run crap-free and full speed again as well, by cleaning out all the obscure registry entries and system services and automatic updaters and cached thingies and temporary wotsits. You just have to know what you're doing.

Also, as with the part of the Linux strategy you forgot to mention, you have to be willing to spend forever doing it, because the tools provided as standard are just about hopeless.

I think maybe you read more into my post than was really there. For example, I never suggested we supply all drinks in pints, only milk and beer, because those are what most people are familiar with. Obviously we could sell, say, a half-litre of milk instead of a pint, but what benefit would that actually bring? Every child in the UK grows up knowing how much a pint of milk is, and every shop sells milk in pints, so changing units (and, realistically, slightly changing the familiar volumes as well) seems like a solution in search of a problem to me.

Except that in this country, almost literally everyone uses that system, and what the metricists are arguing for is the "standard" that almost literally no-one uses.

Losing the Mars mission was very unfortunate, but not nearly as unfortunate as seeing, say, an extra hundred people dying on the roads the year after speed limits changed.

I wonder, do you think we self-absorbed holdouts should drive on the right as well?

But that's exactly the point: we are used to ordering a 1/4lb burger or an 8oz steak, so everyone knows what they're getting, so there is no problem. Rather like programming, using the same style as everyone else for what you're looking at right now is more practically useful than trying to enforce universal consistency for all plausibly related things everywhere.

On the evidence so far: No, not really. That's basically my point.

The important thing with units is standardisation so everyone understands the same quantity to have the same meaning. It turns out that engineers are quite capable of using high-precision SI-denominated measurements at work and still going for a pint with their colleagues at the end of the day.

I live in the UK.

Moreover, I know someone who basically spends their day going around pubs ordering pints and then dealing appropriately with the ones who underfill. Who says government jobs always suck? :-)

It's not about work and effort. There have been fatal air accidents as a result of two readings with plausibly similar numbers but very different meanings being shown interchangeably in the same place on aircraft flight controls. If highly trained professional pilots can make that mistake under pressure, then for sure an average driver can.

Moreover, given that there is always pressure to increase speed limits here to 80mph, while on some major roads there is currently a 50mph (approximately 80kph) limit today, there is at least one obvious case where this could go horribly wrong in practice.

Of course not. I'm just demonstrating the hypocrisy of the argument. A lot of people in this discussion aren't really arguing that we should all use SI/metric units across the board, they're just saying they want everyone else to use them when they do.

So do we, when we buy soft drinks. But in my country, we order beer as a half-pint or a pint, and everyone knows what they're getting. Are you suggesting not only that we should change our units to fit your preference but also that every drinking establishment in the country should buy a complete new set of glassware that will hold different volumes that are more convenient in the new units as well and presumably that everyone's prices should slightly change to match?

So, we've got:

1. Speeds (mph) and fuel (mpg in an X gallon tank)

2. Lumber (2x4)

Let's add a few more:

3. Milk (pints)

4. Beer (pints)

5. Ingredients in menus (pounds and ounces)

6. Human weight (stone, pounds and ounces)

7. Human height (feet and inches)

8. Vehicle heights for bridge clearances etc. (feet and inches)

9. Time (hours, minutes and seconds)

10. Date (days, months, years)

And that was just stream of consciousness, without a pause to think of other examples.

Seriously, standards are great. They help us to communicate unambiguously. And we have standards in the UK, and they are what I just listed. No-one here goes to the supermarket to buy 227g of cheese and 1.14L of milk. No-one goes to a car showroom and asks whether the fuel economy around town is better than 7.84L/100km, and most people's instinct would be that a higher number was better even if they had that reference point. A few people might describe their height in metres, but most people would say something like "five foot nine".

For projects where international collaboration is required, sure, agree a standard up-front, and it might as well be SI. Likewise for scientific and engineering applications, everyone is a professional and can agree to use SI. But for day to day life? You'd better hope someone going to a supermarket or a pub knows the same units as everyone else, because asking for 0.28L of beer at a crowded bar isn't going to make you any friends.

I have LibreOffice installed on one of my systems, and it has always been unhelpful about how it works with the task bar on Windows 7. I don't know what they're doing wrong, but nothing works quite right with either the task bar or jump lists.

LibreOffice is, however, the only one of 20+ pinned applications on the system that has this problem. I don't know whether OpenOffice has the same problem, but if so, I'd say it's an anomaly.

I don't have a strong opinion on the management and practicalities of Linux itself; clearly Linux is already stable enough to run useful software on it, because servers all over the world are doing it today. But any operating system, no matter how good, has little value unless there is software to run on it. Right now, you simply can't buy a lot of serious professional software to run on Linux, and the open source equivalents to things like Excel and Photoshop don't cut it.

They appointed Cloud Guy to run the show, at a time when Cloud was a buzzword. No big surprise there from a trendy board/investor point of view, but to anyone with technical chops that move went against basically every major strength Microsoft had left and played straight to their weaknesses.

Based on historical trends, I suspect MS get 2-3 disasters with Nadella at the top before he gets forced out. The difference this time is that now Microsoft itself can probably only survive 2-3 more disasters on the Vista/Win8 scale before it ceases to be a major player in the industry at all.

The worrying thing is that there is no clear successor, with neither Linux nor OS X having the application base to be comprehensive competitors to desktop Windows yet, while the average web app is still a child's toy in comparison to serious software (and often a child's toy with serious security and privacy concerns). It is possible that the 2010s will be remembered as the decade when progress in software development reversed and the industry became dominated by cheap, "good enough" software that left professional/power users out in the cold, though I have some hope that OS X and the relatively polished, diverse and sometimes disruptive applications running on it will take over before all is lost.

Why? I have a large screen and have literally every application I use on a regular basis pinned, as well as Explorer with the directories I most often want to open. For me, the task bar and jump lists were the two UI developments that made Windows 7 a significant win over XP. Most days I don't even open the Start menu except, ironically, to shut Windows down at the end of the day.