A thing that irregularly irritates me about my brain is that it appears to have different algorithms for answering the questions ‘where should I look for X?’ and ‘where should I put X so that I'll know where to find it next time?’.

The usual illustration of this is that I lose something, and after searching everywhere decide I'd better give up looking and buy a new one. When I get the new one, I try to think of somewhere good to put it so I'll be sure of being able to find it next time – and the place I decide on is often somewhere I hadn't yet looked for the old one, and guess what I find there when I do?

My brain doesn't seem to want to make it easy to solve this problem, because its where-should-I-put-this algorithm doesn't like being invoked counterfactually. If I pretend I've just bought a new one and ask myself where I should put it, I'm likely to get the same set of answers as I did from the where-should-I-look-for-it algorithm; it's only when I've really just bought a new one that the real where-should-I-put-it algorithm activates and I'm able to frustratingly find the old one after all.

It's terribly annoying and I wish it would stop it. Bah.

My electricity supplier, on the bills they send me, annotate every meter reading as ‘read by you’ or ‘read by us’, presumably so that if the readings are ever non-monotonic or otherwise unbelievable they know which ones to question.

Last week they stuck a card through my door telling me that they'd tried to read my meter but I hadn't been in (since, just like every other company in the multiverse that visits people at home in the course of its business, they invariably attempt to do so at the most likely times of day for the occupants to be out at work). My previous supplier's suggested fallback procedure in this situation was for me to read the meter myself and give them the readings either by phone or through their website. This lot prefer me to fill in the meter readings on the card itself, and to leave the card in my window so they can see it when they call back.

So I did that, and today I got the resulting bill. The meter readings are exactly the ones I wrote on the card, but they're annotated ‘read by us’! I wonder if that meter reading is now tagged as trustworthy in their database despite them not having laid eyes on the actual meter at any point.

Vampires don't have reflections; everybody knows that. But how do they not have reflections? What's the mechanism? And what useful applications are there for that mechanism?

I think it goes like this. When a photon strikes a vampire, it separates into two different types of semiphoton, which I'm going to call an R-photon and an S-photon. The R-photon continues along the original photon's path, as if the vampire hadn't been there; R-photons disappear if they touch a sensor (such as an eye, or a camera), but if they touch a reflective object they reflect off it and turn back into ordinary photons. (Most noticeable with a mirror, but I think it applies to diffuse reflection off ordinary surfaces too.) The S-photon, meanwhile, reflects off the vampire as if it had been an ordinary body; S-photons disappear if they touch a reflective surface (again, ordinary diffuse surfaces as well as mirrors), but things like the photoelectric effect work just fine, so S-photons interact with sensors exactly like normal photons do.

Hence, if you're in the room with a vampire, you see it by means of the S-photons scattering off it; but if you look in the mirror, you don't see the reflections of the S-photons, but only those of the R-photons which went straight through the vampire.

There are a couple of fiddly corner cases: what happens if the first object struck by an R- or an S-photon is another vampire? After drawing a thought-experiment diagram or two on paper, I conclude that R-photons go straight through additional vampires and remain R-photons only, while S-photons bounce off vampires and remain S-photons only. Any other behaviour would sometimes enable you to see a vampire in a mirror, by putting another vampire either between the first one and the light or between the first one and the mirror.

So now we have a sound theoretical basis for the optical behaviour of vampires, what are its applications? Well, distressingly, no application I've been able to think of quite works.

You'd like to be able to use a wide flat vampire as a one-way wall: angle a mirror at the wall and you can see through the vampire as if it wasn't there, but people on the other side don't know you're watching. Except that doesn't work if they have mirrors too, and in a world where people used this as a means of covert surveillance, anyone remotely paranoid would have a small mirror about their person at all times.

A vampire itself would have the useful ability to wear a pair of glasses with one lens mirrored on the inside, and thus be able to see people sneaking up behind it with stakes. Unfortunately, a vampire so equipped would no longer be entirely invisible in mirrors – there'd be a mirrored lens hovering in mid-air. I'm sure that would turn out to be inconvenient.

A superficially impressive military application requires a vampire which itself has a mirror finish. If you take a laser beam and reflect it off a vampire, you split it into an R-laser and an S-laser. The R-laser is useless as far as I can see, but the S-laser has definite possibilities: it has the same effect on a target as ordinary laser weapons, but without the risk of mirror-carrying enemies reflecting it back at you. Except that even that doesn't work, because vampires reflect S-photons; so if the enemy also has a mirror-polished vampire, they can use that to send your beam back at you. Arrgh!

You'd also have to avoid using a laser working in the ultraviolet frequency range, or else your vampire would develop a completely different optical problem…

(Thanks to several people in the pub for helping me discuss this very silly idea :-)

I fell asleep last night and dreamed the most fabulously complicated heist movie. Together with a couple of confederates, I had visited Russia in the Soviet era and begun to put into motion an unbelievably intricate and fragile plan to steal state secrets from KGB headquarters. (Under normal circumstances that'd have made it a spy movie, but for some reason my dream cortex decided to do it in heist-movie style instead.) There was a lot of stuff along the lines of making multiple visits to the HQ disguised as different workmen, and drilling holes in just the right places so that eventually there'd be a bunch of holes in walls lined up in a perfectly straight line reaching from a sensitive interior room to the outdoors.

The preparation stage had been going on for hours of dream-time and was nearly finished … when, to my utter frustration, I woke up, and was unable to resume the dream when I went back to sleep. Arrgh! I really wanted to know what those lined-up holes (and all the other similarly complex bits and pieces) were going to be used for. Now I feel as if I wasted all that time.

I drove down to London yesterday to visit my sister, who lives in the general vicinity of Wandsworth. Normally I'd use a more sensible mode of transport, but on this occasion the point of the trip was to bring some bulky stuff back to Cambridge for Mum to look after while Sophie's away in New Zealand.

I hate driving in London. Every time I do it, I vow never again, and someone always manages to talk me into it regardless. Still, I might as well keep trying. Never again!

When I first started driving, I had an absolute terror of London, because I'd heard so many awful things about London drivers. Driving around the M25 was fine, but I had a lifetime ambition never to take my car inside the region it enclosed if I could possibly help it.

I think I first broke that ambition by accident, by missing my turning off the M11. That was OK; I just turned round at the first opportunity and hastily went back, and it was still just ordinary motorway and roundabouts. Then I visited somebody only just inside the M25 on the north side (Enfield or thereabouts, if I remember rightly), and that didn't seem obviously scarier than outside.

With my confidence thus boosted, I made a couple of trips to places on or around the North Circular, and then one to the Isle of Dogs. That was where it started to get irritating, not because of the feared London drivers but because the road layout policy in London is alien to all my non-London experience and I can never predict what the lane system is going to do next. The last time I went, I had drswirly in the passenger seat being a full-time navigator and we still managed to take several wrong turnings and end up both thoroughly infuriated with the wretched city's road designers. ‘Let's never come here again,’ said Gareth. If only. Still, it seemed like an improvement to avoid London out of annoyance with the road layout than from terror of the danger to life and limb from the sinister phantom of the London Driver.

So yesterday I ventured into Wandsworth by way of the M4 and the South Circular. On the way in, again, I had a full-time navigator in the passenger seat, and apart from an easily corrected wrong turn in a one-way system five minutes from our destination, we got the whole journey done with only a constant background level of annoyance at the lane system: the road kept dividing into two lanes and back again for no very clear reason.

On the way out again, my navigator was missing: Mum didn't need to be back in Cambridge as early as I did, so she'd decided to stay a bit longer and take the train home, leaving me to drive all the bulky luggage home on my own. I wasn't too worried by this: getting out of big cities is always easier than getting in, because there are always lots of signs pointing to the motorways.

It went fine until I was nearly back to the M4, at which point I took a wrong turn at a roundabout because the right turning had no confirming sign. In retrospect, actually, it did: it had a sign after the turning, pointing back the way I'd come saying ‘this is where you should have gone just now’. The spiralling lanes didn't permit me to go all the way round and try again, so I ended up heading off in the wrong direction, and then decided to take my life in my hands by trying to find my way back to that roundabout without stopping to look at a map.

It worked, but only just. I turned off the random A-road I was on, and attempted to head back in the right general direction by counting my turns and hoping to hell all the turns were good enough approximations to right angles. Just as I'd lost count, I found myself at a crossroads in the middle of Chiswick which apparently hadn't heard of the idea that signposting the major motorways is helpful. Fortunately, I noticed a little brown sign pointing to ‘Brentford Fountain Leisure Centre’, and in a feat of memory which I'm still feeling smug about the next day I recognised that as the name printed on a building we'd driven quickly past on the way in to London five hours earlier. So I headed in that direction, and was right to do so; within minutes I saw the roundabout ahead, took the right exit this time, and was on my way home with no further mishap.

So I was pretty pleased with myself for managing to get myself out of that, but pretty irritated with London for requiring me to have that good a memory!

Also, this time, the real London drivers finally did make an appearance. I think they must have observed the same indecisive lane markings as me and drawn the conclusion that they were probably only advisory; so they kept treating ordinary one-lane-each-way roads as if they had two lanes, and overtaking me on the inside at moments that had me shaking my head and wondering how anyone who did that on a regular basis was still alive. Fortunately my driving experience to date had left me with good instincts for spotting someone about to undertake me and a good reflex response to start behaving especially predictably as soon as I notice anyone else doing something weird, and those kept me out of serious danger. I still felt like a herbivore among apex predators, though, and was very glad to get out of there.

Let's try this one more time. Never again!

It is a lifetime ambition of mine that one day I may hear someone utter the phrase ‘It's cold outside’ and not immediately have my brain start playing the Red Dwarf theme to itself on repeat.

There's a large class of global optimisation algorithms which share a common dynamic-programming sort of approach.

( this one is serious geek material )

So because these things keep coming up, and in particular because I keep finding applications of the same principle to solve problems I'm faced with, I would like there to be a piece of terminology that precisely describes this particular optimisation strategy.

‘Dynamic programming’ is of course an umbrella term which covers all of the above. But it's too general: it also describes other types of algorithm which don't fit into this specific framework. I want a word for this specific shape of dynamic programming algorithm. The best I currently have is ‘Viterbi-like’, and that's useless because even I wouldn't have known what it meant until a couple of years ago.

Suppose you have a thing which you're comparing to another thing. (Whatever they might be. Consumer products, pieces of software, business models, algorithms, I don't care.) Suppose there are a number of criteria on which you might compare the two things, so that there are two ways in which the comparison might be inconclusive: as well as ‘they're both the same’ the answer might be ‘better in some ways, worse in others’.

The thing I keep finding I need a word for is the situation where neither of these is the case: where one of the things is better in at least some ways, and although they might be exactly tied in other ways there is no way in which it is worse. This is the point at which it typically becomes a no-brainer to throw away the other thing and adopt the better one, whereas in any other situation you might hesitate for fear that one of the ways in which the new thing is worse might turn out to be the most important criterion.

I've heard people use – and found myself unconsciously using – a lot of different words for this, but none ever seems quite right. ‘Uniformly superior’ isn't right, because often it's not actually superior in every single way: merely superior in some and equivalent in others. ‘Linearly superior’ is one I've heard a surprising number of times, and it always seems to make sense in context, but when you look at it more carefully there isn't the remotest connection between this concept and any of the usual meanings of ‘linear’. ‘Unconditionally superior’ is one of the better ones, suggesting that its superiority is not conditional on the relative importance to you of the various criteria, but again it has a bit of the ‘uniformly’ problem, in that if one doesn't pay attention it's easy to read it as suggesting that the thing is actually better in all ways.

In mathematics, there is a precise term which means what I want: ‘greater in the product order’. (A product order is one possible way of combining many individually comparable quantities to produce an overall comparison of the lot, and it states that one list of quantities is greater or equal to another list if and only if each individual quantity in the first list is greater or equal than its counterpart in the second. So, ‘greater or equal in the product order’ means that the thing is at least as good on every criterion, and ruling out the ‘or equal’ clause means that there's at least one criterion on which it's actually better.) However, on the rare occasions that I've tried using this phrase for this purpose it has confused even other mathematically trained people.

I'm sure there ought to be a sensible and widely understood phrase for this concept, because I find myself needing to use it so often.

Pedant points:

• Use common sense as necessary for borderline or ambiguous cases
• Use your own judgment as to what reasonably counts as ‘often’
• By ‘music-playing equipment’ I mean equipment for playing recorded music; musical instruments are outside the scope of this poll
• Devices which can play music but are not specifically intended for so doing (e.g. a computer that just happens to have a sound card) should be considered music-playing devices if you ever play music on them; otherwise make your own judgment.

My plan to buy a Chumby and convert it into a somewhat over-engineered alarm clock has now come to full fruition, and appears to have been entirely successful.

I had to junk the original version of the software I wrote for it, because it turned out that the Chumby's Flash player was apparently never intended to keep running the same program for more than a few minutes at a time: every three days or so it would leak enough memory that the clock became unresponsive and had to be rebooted. So I threw the whole thing away, thus fulfilling Brooks' Nth Law, and wrote it again from scratch in C. It is now lovely and fast, it doesn't leak, and as an added bonus it's much easier to persuade it to store persistent data and make network connections.

So the clock program now does everything I originally envisaged, including tying in over the network to my existing calendar software. Yesterday was the first test of this on live data: it correctly spotted the impending bank holiday and automatically cancelled the usual weekday alarm arrangements, and it has correctly not tried to do the same thing again for tonight. Success!

So now it's working, the real question is: was it worth it? And already I'm confident that the answer is yes.

As I originally planned, I can now get up early with impunity. I can just press ALARM OFF, get out of bed, and not worry about it. Where I would previously have had to come back to the clock at 8am to stop it beeping, or remember to turn the alarm back on properly that night, I now need do nothing at all – and on the next school night, it's lit up just as it usually would be, reminding me to enable the alarm for the next morning. One irritating little weight off my mind.

But there's an additional advantage which I hadn't considered, which revolves around the snooze function. With my previous clock's snooze function, you would press the Snooze button and then be pretty much committed (on pain of thundering inconvenience if you go back on it) to staying in bed for another nine minutes. So you might as well drift back off to sleep, and then when the alarm goes bwarp again nine minutes later you're just as unwilling to get up as you were the first time.

But with my snooze function, because you can trivially cancel it at any time in the middle of the snooze interval, it's now feasible to lie in bed and gradually wake up rather than going straight back to sleep; my normal usage of the snooze function is now to hit Snooze, lie around for three or four minutes, then cancel the snooze and get up. The alarm at the end of the snooze period has now become a safety feature just in case I accidentally drift back off during this process; in normal usage it never actually goes off. It's as if my redesign has completely changed the nature of what the snooze function is fundamentally for, and the new model feels much more sensible and useful. (Of course the old usage is still available if I should want it: there's nothing stopping me hitting the snooze button and dozing straight back off).

There's always a danger, when you want something for ages and finally get it, that it will turn out not to be as good as you'd hoped, or that your imagination had not correctly envisaged the way it would be used and that it will therefore turn out not to be as useful as you had believed. Today I own an alarm clock which has well and truly avoided these traps, by not only fulfilling all my original expectations but also by turning out to be useful in extra ways. I got it right!

Rather to my own surprise and several other people's, I went to the beer festival last night.

This was the first one I'd been to since 2004, and not surprisingly since coeliac disease renders me permanently incapable of drinking (normal) beer, so one would naturally expect that beer festivals were things to which it was pointless for me to turn up. I was persuaded to come along anyway by a variety of lovely people, and I spent the evening drinking cider, which wasn't bad.

I've often said that the thing which annoys me most about cider is that very tart dry cider seems to be in the overwhelming majority, whereas my taste runs more to sweet cider. (And, to be honest, sweet alcohol in general; for example I'll always pick sweet sherry over dry even if the former is cooking-grade and the latter really poncy, and I'm a great fan of mead.) This was illustrated particularly graphically at the cider bar last night, where they had about thirty different ciders available of which about five were sweet – and yet, when I went back to the bar for my second pint, all the sweet ones had sold out first and I had to make do with medium-sweet. What's with that? Everyone appears to want to drink sweet cider, but nobody seems interested in making it. Aren't market forces supposed to sort this kind of thing out?

Two people went out of their way to tell me that before I arrived the Tannoy had announced the availability of a gluten-free beer. So naturally I went in search of that at one point. I couldn't find it, and eventually resorted to asking a staff member – who turned out to have just spent an hour searching for it on behalf of another drinker, without success. So I didn't get to try that, which was a shame.

My usual irritations with the beer festival as a drinking venue were somewhat mitigated by the fact that for a change it didn't rain. In fact I think this might have been the only time I've ever been to a beer festival and not had it rain us all into the inadequately sized tent. The bar was still a mad scrambling crowd and I still wished I'd brought a chair of some sort, but it could have been worse. And I got to see lots of lovely people and even meet a new face or three, so for socialising sorts of purposes I was definitely glad I'd gone.

As usual with beer festivals, I felt hung over this morning in astonishing disproportion to the number of units of actual alcohol I had. Never quite sure what that's about.

( why programs that do nothing can come in handy )

So nullfilter, despite my having originally assumed it would be a completely useless program for actually using, has proved its worth repeatedly in the last month and in fact might well end up being the one of my filter programs from which I get the most ongoing use!

I almost feel there ought to be a moral here. Perhaps it should be a comment I recall once seeing posted on a newsgroup by (I think) pm215: ‘never neglect the trivial case’.

People conflating practical fixability with blame. Suppose something bad happens, and the person who is morally speaking most blameworthy is someone who it is practically speaking impossible to induce to modify their behaviour, stop doing it, fix the problem, whatever. In that situation you might regretfully have to approach somebody else involved in the problem, whose fault it either isn't at all or is significantly less, explain why the problem can't be solved in the ideal way, and ask them nicely to modify their behaviour instead. What you do not do is to go to that person and rant and rave at them self-righteously as if all the moral fault from the original situation is theirs!

I've just discovered that if you write the following code:

    if (condition);
    {
        statement;
    }

then gcc will accept it without question, whereas Visual C++ will give a warning pointing out that the semicolon after the closing parenthesis was probably not what you actually wanted to do. One point to Visual C++; none to gcc. Unusual, since gcc normally seems better at this kind of ‘probably not what you meant’ warning.

(I wouldn't have made the error at all if it hadn't been for this wretched code I'm maintaining which was written by someone who didn't think there was anything wrong with five-mile-long individual source lines. I can't reliably check there isn't rubbish at the ends of the lines if I can't reliably see the ends of the lines!)

For any interested geeks in the audience, here's some more technical detail on the Chumby and my use of it as a high-tech alarm clock.

( geeky geek geek )

So, about a month ago, I complained in here about my alarm clock having particularly poor behaviour with respect to DST changes. A few days later I was ranting to some friends about this, and tangentially about how alarm clocks always irritate me anyway because their user interfaces never quite do what I want, and for decades what I've really wanted in this area has been an alarm clock for which I can write the software myself.

My main annoyance has always been how you turn the alarm on and off. On traditional analogue alarm clocks, you manually enable the alarm when you go to bed each evening, and turn it completely off in the morning as a means of shutting it up. This has the obvious failure mode that if you aren't paying attention when you go to bed, you won't get up on time. The standard digital alternative is that you turn the alarm on and then it beeps at the same time every morning until you turn it off again; that has the slightly less obvious failure mode that if you're unexpectedly elsewhere it goes off anyway and irritates anyone nearby, and also that if you wake up early and decide to get out of bed, you still have to come back and stop the alarm beeping, or alternatively shut it off completely and (again) have to remember to switch it on again that night.

So my solution involves an alarm clock with three basic states: alarm on, alarm off, and ‘you haven't told me yet’. In the last state, there's a big visible indication of some sort on the display which you'd have to be extremely absent-minded (or drunk) to miss. So essentially this is the analogue model – you have to manually enable the alarm every evening that you want it to go off the following morning – but with a big reminder making it nearly impossible to forget to do so. This seems to me to manage the best of both worlds.

(In addition to that, if I was writing the software myself anyway, I wanted a number of other minor tweaks; for example, the ability to set a one-off different alarm time and have it automatically reset to the standard one the next day. Or the ability to use the snooze function as a countdown timer at any time of day instead of only being able to activate it when the alarm is actually in mid-beep. But the tri-state thing was by far the most important point.)

So I ranted this at post-pizza a few weeks ago; this turned out to set in motion a chain of events which involved me and Ian getting about half way through designing an alarm-clock-shaped custom computer peripheral before someone pointed out that a Chumby might be just what I wanted: http://www.chumby.com/. Essentially, it's a small computer with a touchscreen, in an alarm-clock-sized package, able to synchronise to the correct time over a network, and (relatively) conveniently programmable by the end user.

The Chumby Store won't ship outside the USA, but with a little help from my friends I was able to arrange for one to be posted to me anyway. It arrived two days ago, and since I'd been able to write most of the software in advance, it is now sitting on my bedside table acting as a nearly ideal alarm clock, and successfully woke me up this morning. Hooray!

(I say ‘nearly’ ideal because there are still a couple of annoyances, most notably that the sound output is glitching in a really annoying way and I haven't yet worked out why. I'm sure it's in my software, because everyone else's Chumby applications sound fine. But as ewx pointed out yesterday, having your alarm clock make a really annoying noise is not necessarily a bad thing!)

Yesterday I changed my work password. Today I got in to work, sat down at my computer, and didn't absentmindedly type the old password. A good omen for the day, I felt.

Two minutes later I was forcibly rebooting said computer in the hope that when it came back up it would talk to the network file servers more usefully. So much for good omens.

My colleague Peter just produced this excellent bon mot:

C++ shares with Lisp the property that you shouldn't be allowed to use it until you've used it for two years.

(He says he can't remember having seen it elsewhere, although of course people do occasionally forget that sort of thing. Whoever said it first, though, it's very good :-)

Probably most of my friends have heard me waffling on about the Infinity Machine at one time or another. If anyone reading this has managed to miss it so far, you can find my article introducing the concept at http://www.chiark.greenend.org.uk/~sgtatham/infinity.html. Today I have a question about it to ask geek-inclined readers.

The Infinity Machine has been a fun thought experiment for most of my life, but in one respect it's slightly frustrating. Its ability to search a (countably) infinite space in finite time would enable it to solve quite a few problems that are famously unsolved in the real world; but quite a few of those problems would simultaneously be rendered pointless to solve anyway by the presence of Infinity Machines in the world. For example, you could use the Infinity Machine to search all possible computer programs to find the one which was fastest at factorising large integers – but you wouldn't want to, because if we had Infinity Machines then a perfectly naïve factorisation algorithm would be just as efficient in practice and far easier to get working correctly.

It occurred to me this week that there is a scenario in which that slight frustration might be resolved. Suppose you were suddenly taken away from your normal life and sat down in front of an Infinity Machine for a few days, or a week, or a month. Suppose you were free to write programs and run them on it, and free to write finite quantities of the output to real-world storage media to take back with you, but when your time expired the Infinity Machine would vanish and nobody would ever get their hands on one again. In this situation, asking the Machine for efficient finite algorithms would be entirely sensible, in principle.

What would you get it to compute, in this situation?

(Ground rules: to help you write your programs you can have access to a large library, and perhaps archives of reference websites such as Wikipedia if you want them. But you don't get unfettered Internet access while you're using the Infinity Machine: I don't want you doing things like factorising every key on the PGP keyservers, or all the root CA keys, because that's against the spirit of what I'm interested in asking.)

Some thoughts on my own answer:

( you might want to think of your answer first, though I don't insist on it )

So, what do other people think?

Perhaps I should ask some subquestions as well:

• What could you do that would have the biggest effect on improving the world?
• What would you do in order to make yourself the biggest profit?
• What would you be most curious to know the answers to, even if nobody would ever believe you?
• And which of those would you prioritise most highly: what would be the thing you'd actually do if given the chance?

Misunderstanding what the difficult part of a job is. Specifically, the misapprehension that the hard part of a given job is thinking of a starting point or basic approach to the problem at all, when in fact the hard part is getting all the implementation details to work right given such an approach. People under this misapprehension will cheerfully suggest a variety of starting points, some of which might even be the right one but certainly none will be a good one you hadn't already thought of; they will then be puzzled when you look more rather than less irritated, and mysteriously don't thank them profusely for their vital contribution and spring immediately into action.

In extreme cases of this, the person will provide several basic approaches and (implicitly or explicitly) suggest that you ‘simply’ try all of them and see which one works best, apparently blissfully unaware that they've just attempted to multiply your workload by three or four (or perhaps even succeeded in doing so, if they're your manager).