Building 26 is infested with singing mad scientists whose accents are almost as scary as their hair, trying to determine whether toenails qualify as the human tissue they need to complete their evil master plan.

Incidentally: for those who've seen District 9, isn't it an awesome movie? For everyone else: don't see it. If you haven't already, you haven't the stomach for it. I got through the messy bits by virtue of being fortuitously surrounded by a mob of stalwart MITSFSians.

120 tubas of all shapes and sizes, including several double-bell euphoniums and a battery of five matching silver sousaphones, performing Christmas carols outside Faneuil Hall: a chilly but cheerful experience.

Fantastic Mr. Fox has some good advice on what not to do about midlife crises, how not to purchase real-estate, and how not to raise well-adjusted children. Also a brilliant send-up of cricket. I'll have to watch it again a few times to catch all the jokes I missed while I was blinking.

Normally I'm fairly disciplined when gift-shopping, but when The Complete Saki wandered into my hand on the weekend I succumbed and took it home with me. My mother has a collection of his, including possibly my favourite short story ever, The Unrest Cure. Saki had a wit somewhat like Oscar Wilde's ("Major Pallaby was a victim of circumstances, over which he had no control, and of his temper, over which he had very little", "His baptismal register spoke of him pessimistically as John Henry, but he had left that behind with the other maladies of infancy, and his friends knew him under the front-name of Adrian.") but a much darker outlook (babies sporadically get eaten by wild animals when things look like they might get too cheerful). It'll keep company with my volume of Dorothy Parker.

Resuscitating desktops whose original operators are long gone (taking their passwords with them) is a good demonstration of the vulnerability of any machine to whose power switch an attacker has physical access. A few minutes spent with ophcrack takes care of most Windows passwords. *NIX and OS X users don't get to be smug about this: rebooting into single-user mode and erasing the root password is just as efficient.

Doing dead code elimination by hand in data-flow languages is easy (look for a wire that goes to nowhere; remove it; repeat as necessary), until you let loops with state into the language. That's when the appeal of automatic theorem provers becomes much more obvious.

The incidental music to A Midsummer Night's Dream makes much more sense when you have a play happening around it. MIT's symphony orchestra and theatre program combined to give a spirited if slightly rough-edged performance of the two together, to good effect.

Someone has put an elephant on my desk. All the usual suspects have been questioned and ruled out. I am perplexed.

View your graduate career as education, not training. Graduate school trains you for precisely one job, and your adviser already has it.

With any luck, though, it also educates you well enough that you can do something else afterwards.

O'Flaherty said, "When I was little, I remember how astonished and interested I was in how easy it is to take life—when you're driving a car, all you have to do is move the wheel a few inches to the left and you kill somebody and also die yourself—and how difficult it is to keep alive someone who is sick. Building is interesting, because it's ultimately impossible, I suppose, but killing is boring. It's easy to see through something—to see how stupid it is, or how wrong—but that doesn't take very long, and then you're finished. If you want to show how boring India is, and how stupid it is, and how wrong Freud was, and how stupid analysts are, that doesn't take more than a year or two. [...] Killing may be amusing while it lasts, but it never lasts very long, and you are back where you started. Killing doesn't solve the problem of boredom"
―Janet Malcolm, In the Freud Archives

UPDATE: Taken out of context, it is not clear that this passage is metaphorical. "Killing" is to be understood not as the literal taking of life but as destructive activity in general: attacking, belittling and demolishing the work of others is quick and easy, and leads some people to sense of jaded superiority. Constructive activity; the building up of ideas and structures that were previously absent, is a much more challenging endeavour whose rewards come much more slowly. It's easier and more fun to knock down a cathedral than to build it, to criticize than to create; does that make the demolitions expert and the critic superior to the architect and the playwright? Which would you rather be?

It was an awe-inspiring spectacle. On one side, a thin line of bright, bushy, but green volunteers commanded by a handful of sleep-deprived but experienced admins; on the other a flood-tide of 2700 high-school students and attendant relatives [1]. Splash has gotten big. The program is now limited by the available classroom space at the institute, and most of the courses are booked solid [2], often within minutes of the catalog being posted online. So far everything has kept functioning with only minor hiccups, but if demand for the program keeps growing at the current rate I worry it might become a victim of its own success. Everything looks straightforward when you're a teacher, but helping out with the administrative side makes clear what a huge mobilisation of forces is required to make the weekend happen and how thinly the volunteers are stretched. It's a good lesson in following orders.

The teaching part of the day went well enough. My Lorentz equation class went according to plan, covering everything I wanted to cover but without much sparkle. The limits class still needs some work, as I haven't adequately defined the objective of the class. Still, I waved my arms about enthusiastically, dangled shiny facts in front of impressionable young eyes, and generally put on a sufficiently good show that the students got involved and interested anyway. I was much happier with the level of participation in that class even if the material wasn't yet as well-organised as I'd like it to be.

[1] Large concentrations of mothers each convinced that their boy is more uniquely gifted and special than all the other keen kids around them are unstable. Disappointingly, I didn't see any mothers convinced that their daughters were uniquely gifted and special.

[2] I spent much of the day at the scheduling and class-changes desk, trying to convince people with unfilled slots on their schedule that they really wanted to hear about Quaternion Algebra or the Sociology of Disney Princesses, those being the two classes that had a few spots left.

Shameless plug: for a heavily abridged and pedagogically reordered summary of what I've been up to for the last year, you can see the (unpublished, unrefereed, subject to modification or improvement etc...) preprints for a theory proposal and experimental demonstration on the arXiv.

I had a discussion yesterday about the desirability of requiring disclosure of data collected for published research. Certainly papers should present evidence adequate to support their conclusions, and it is part of the referees' job to ensure that. However, that evidence is often presented in highly summarised (often graphical) form, and some people would like to compel researchers to make all the raw data sets publicly available as well, instead of merely keeping thorough records that can be checked if a difficulty arises. There are objections to this superficially laudable policy:

• A data set often represents years of work for one or more graduate students. The principal reward for this work is the right to author the subsequent papers, build whatever reputation one can from them, and earn a shot at a career. Once the data is published, all the teams in the field with nothing better to do can process it half-a-dozen different ways and get all the juice out of it, leaving the students who took it high and dry. Of course, data analysis might speed up as a consequence, but data production would slow dramatically: lots of good scientists will work for minimum wage if you dazzle them a bit with the hope of making a name for themselves; hiring them without that intangible incentive is a much more expensive proposition.
• The data released should be at the right level of abstraction to be useful to the community. In some communities, the apparatus and procedures are standardized; the language of those communities allows the efficient communication of experimental protocols and raw measurement results, and those are then a productive thing to share. In some communities, the systems being studied are much better understood and standardized than the instruments used to study them. Documenting the raw datasets in a form comprehensible to outsiders is then a waste of effort because they're at the wrong level of abstraction anyway. The information needed to reproduce your results is not the output of your particular non-standard instruments, it's a description of what you did to your universe-standard atoms.
• In communities that have attracted the attention of conspiracy theorists and cranks, I can imagine feeling a strong reluctance to placing one's data (and hence one's name) at the disposal of all reinterpreters. I don't think this is really a valid objection, but given the value of a scientific reputation I can at least sympathize with a desire to prevent the barbarians at the gate from turning it against itself.

Mandatory data set disclosure is a lovely concept in the idealized scientific method we learned about in high school, much as communism is a wonderful concept in the idealized human society we were taught about in elementary school. It probably even makes practical sense in certain cases. But I'd be careful about painting its opponents black.

Fundamentally, you can't mandate scientific honesty. Any sufficiently determined liar can fool a community for a while [1]. Fortunately, science is all about weeding out errors, be they honest or willful. It's just a matter of time and a preponderance of truthful [2] scientists. Since a disregard for the truth opens up a variety of more rewarding career paths, I'm broadly optimistic that the latter condition will be fulfilled most of the time.

[1] All you need is a function

mysignal(x) = myfavoritemodel(x) + plausiblenoise(x)*prng()

in your favorite programming language and the rest is all story-telling skills.

[2] Note I said "truthful", not "kind", "pleasant", or "politically trustworthy". Those are nice too, but hardly necessary for present purposes.

I borrowed from the library a volume of "Experimental Wireless and the Wireless Engineer" from 1930, tied together with cloth ribbon. While the style of the writing is as old as the paper on which it is printed, the contents are quite similar to those of the latest issue of Wired: trade show reports, illustrated reviews of new gadgets, patents, and textbooks that are up-to-date with the new electron theory, as well as discussion of practical methods for stereoscopic (i.e. 3-D) television broadcasting. In fact I'd say that about half the articles in the volume related to television technology in some form or other. As a reminder, this was written when steam locomotives were still the normal method of long-haul overland transport.

The halting theorem as Dr. Seuss would prove it.

The recent outbreak of hysteria over the new Google-sponsored Go programming language (which I first found out about when I followed a link I believed to be about computational structures in ancient Far-Eastern strategy games) has reminded me of a nagging feeling I've had for a while now. For all that they spent decades berating each other and using each other as examples of what they objected to, the Bell labs languages (C, Plan9 C, Aleph, Limbo, and now Go) and the Wirth languages (Pascal, the Modulas, the Oberons) have been walking much the same path, from much the same prejudices and with much the same results, for most of that time. They're designed by systems-builders with a deep appreciation for simplicity and a talent for finding clean ways of eliminating problems rather than complicated ways of overcoming them. Both are biased towards making life easy for the compiler-writer and for building the bottom layer of computer systems, because they were designed by people who built their computer systems from the ground up. Both are often marred by strange limitations, elisions, or special privileges for the built-in library that make the compilers easier to write. Neither has strayed very far from the programming models of Djikstra and Hoare; whether because they ignored all the cool programming languages research of the past three or four decades or because they decided they didn't trust it. In fact the only fundamental distinction I see is between a taste for funny punctuation characters on the one hand and obnoxious capitalization on the other. Ironically, Go's export of capitalized identifiers is taken straight from Oberon, which exported by prefixed punctuation mark.

Someone has found a new corollary to Parkinson's Law! The complexity of the tax code, pricing model, or other bureaucracy to be implemented will grow as much as computer automation allows. Consequently, automating bureaucracies is not only ineffective; it is actively harmful.

One Matt Purkeypile has put forth a doctoral dissertation on how to do large-scale object-oriented commercial software development on quantum computers. I suppose I should commend him for thinking ahead, but I feel no urge to do so. For one thing the abstract begins with "While not yet in commercial existence, quantum computers ...", which is one of the more egregious euphemisms I've recently come across. It ends with "... the fundamental concepts that make quantum computing practical for common developers", eliciting horrible visions of Microsoft™ Certified™ Quantum™ Solutions™ Developer&trade (we offer solutions in increments no smaller than h$) and questions on web developer forums about whether you can use your hard drive for extra storage when you run out of Hilbert space. Gah. Solving abstract problems of no real-world import is one perfectly worthwhile thing, meeting practical challenges without making any deep intellectual advances is another, but do we really need to combine the worst of both?

I realize that I mostly say nasty things about quantum computing papers here, partly because the laypeople and theorists mostly seem so googly-eyed about them, but I should end on a positive note. Hence: one good paper discussing how the formal tools of quantum computation theorists can give productive insights even if no quantum computer is ever built, and Peter Shor's original factoring paper has a nice introduction to a useful subset of quantum mechanics for the mathematically astute layperson.

I'm teaching two classes at Splash this year, and both are booked solid (as is the rest of Splash ... scary). One is a repeat of the "Derive the Lorentz Equation using symmetry and middle-school algebra!" class I ran in the winter of 2008. Fortunately and quite accidentally I just found the notes I made for it at the time, and they are ridiculously complete. I'll probably edit them a bit for gender-neutrality and consistency of Greek letters, but otherwise that class should run fine as long as I watch the clock.

For the other class, however, I have two weeks' worth of spare time in which to come up with a plan. It is meant to be an introduction to the type of fundamentalist thinking that makes good scientists and a certain kind of good engineer. Where by fundamentalist thinking I mean thinking about the fundamental limits on processes. So I'm looking for processes whose fundamental limit I can calculate in terms comprehensible to precocious high-schoolers. Here are a few ideas, some less promising than others:

• How far away can you see a mouse? (Calculate from the angular resolution of your eye, then from the angular resolution of a diffraction-limited eye, then from a hypothetical oversized eye ...)
• How fast can you get water to flow through a pipe? (Limited by the onset of cavitation, which you can estimate from Bernoulli's principle; I might sidetrack a bit about Reynolds' number and the breakdown of laminar flow just because it's cool)
• How fast can you fly an aircraft? (Sound barrier? Heat barrier? Speed of light?)
• How tall a mountain can you make? (I gather that an argument based on comparing the gravitational potential to the binding energy of atoms in rock gets Everest about right, though I need to work it out myself.)
• How efficiently can you transmit power over long distances? (Going to higher voltage cuts down on electrical transmission losses, but then you need bigger towers to get the conductors a safe distance apart; 750kV seems to be the current limit, not sure whether that's limited by the economics of large towers or by materials constraints at the endpoints.)
• How accurately can you measure electrical current? (The Johnson noise limit is one of the few measurement limits I know of that is reasonably tidy but not quantum-mechanical ...)

If anyone can suggest other tasks or processes to use as examples, I'd be grateful. Even if you can't identify an interesting and easily-derivable fundamental limit for it, I want to hear about other ideas; I'm in brainstorming mode. Hopefully they'll let me bump some of the weaker ones in the preceding list and still have a full hour-full of fun examples.

Tonight, after longer than I care to think about, M and I hit "Submit" within three feet and two seconds of each other, forwarded each other the confirmation emails, and left. Sending off another two papers to the journal doesn't solve our problems (after all, we're still arguing with the referees about the first one), but it definitely changes them.

So I've had a monster bowl of Phô Pasteur's soup, a mug of Chamomille, and a CD-full of my childhood drinking songs (courtesy of a British Columbian collector who has most of the Irish Rovers' recordings). Time to rest up for the next adventure.

David Mermin's Knight Lecture on "Writing Physics" is spot-on. From the frustration of discovering that your coauthor cares about writing just as much as you do, through the annoyance of going through and wiping out every trace of style from your papers in the names of brevity and objective writing, to the concern that many smart people expend enormous effort attempting to solve pseudo-questions that arise only because of misleading use of natural language.

It's all worth reading, but if you're in a rush skip to the last three or four paragraphs. They clearly state what I've always wanted to believe: that finding the right model and the right words to explain a result clearly is as important as obtaining the result in the first place.

Thanks to nakor, I experienced my first tabletop RPG run this past weekend. It was great fun: everyone was considerate and helpful to the newbie and we saved the world from tentacular devourment. Some reflections:

• The purpose of the exercise is collaborative story creation (with a random-number generator thrown in to stop things getting stale), not tactical problem-solving as such. I partly understood this going in, but hadn't fully grasped the implications. In real life, a dull and uninteresting solution to a problem is still a solution; in-game it's cheating the other players of the more exciting story that might follow if your character failed. That's a bit hard to take for someone with an engineering mindset.
• Pick characters you like, but more importantly pick characters you can make interesting. I missed a few hints and ended up with a British Zorro or Scarlet Pimpernel character; great character, tactically useful, just the sort of hero I like to read about, but difficult for a beginner to write on the fly. His social talents, in particular, were underutilised. In hindsight, I should have taken the part of the Three-Fisted Midget. Even a beginner can write the Three-Fisted Midget amusingly ("a diversion is needed: the Three-Fisted Midget casts the croupier at the bartender and thereby adds an aspect to the scene of 'all NPCs in the joint are staring at the short berzerker in the ill-fitting suit'"), at any given point in time the next move is obvious (hit something), and you can use the extra brain space to observe and learn from those around you.
• A few points about snipers:
  1. Sniping seems a reassuring role for a beginner: you get to hide in dark corners and smite things from afar. It's a power trip for the sniper. It is disappointing to anyone else: "And then the bad guy was shot from afar, obviating the need for the final confrontation." just doesn't work very well.
  2. Sniping won't work the way you expect it to. Since no one has fun when characters drop dead too easily, the rules are set up to make it unlikely that you'll do too much damage in a single shot. Your hope of an elegant (but narratively uninteresting) surgical strike is now replaced with attrition-at-a-distance.
  Don't plan on being a sniper unless your team needs one, and then think of it more as long-range boxing. I did end up firing the parting shot of the game, but thanks to a team effort involving truly spectacular feats of derring-do by other players that cleared away the minions and let two of us keep shooting at the cult leader until he allowed himself to be committed to an asylum for the insane.
• There was much unexploited scope for teamwork in our story, no doubt because it was a single-session game where we hadn't time to become properly familiar with the characters. Perhaps also because the character of the natural leader and organiser was being played by someone who was too busy following and learning. For instance, the brilliant wealthy engineer kept coming up with fascinating widgets to expand his own otherwise modest talents; what if he'd leveraged other characters' skills instead (pheromone-enhanced femme fatale, improved sensory apparatus for the watcher from the shadows ...)? Instead of all blundering around together asking for the MacGuffins, what if we'd set up big diversions and let the ninja sneak around unobstructed? And why did we let our scientist serve as diversion and bendy-toy for the Cthuloid tentacles? It seems a misuse of his talents.
• A lot of things get much funnier with even a few hours hindsight. Hot potato with a stick of dynamite in the midst of a crowd of hostile cultists carrying pails of fish. Repeated use of ballistic Studebaker as a crude ranged weapon. Convincing ourselves that one of our more useful team members was a traitor, and then coming up with excuses not to shoot her in the back for the rest of the game.

I'll go back to dealing with the real world for a while, but I enjoyed this foray into an imaginary one.

I love it when a student I've left to stew in their own problem for five minutes says "oh wait!"

I saw a pure white squirrel this morning. Not an albino, as far as I could tell (the eyes looked normal). He looked scared. I don't blame him.

As I beat a slow but steady retreat before the bug that has felled most of my labmates over the past month, and hope that it doesn't get bad enough that I have to quarantine myself, I've gotten creative.

Chicken soup for breakfast! (Vegetarians can substitute something appropriately sinus-clearing like cayenne squash stew).

Of all my crazy inventions over the years, this one might have the most promising future.

Ever since my first year of undergrad, when I sank countless hours into building sets for SkuleNite, I've kept my extracurricular self in check. A bit of singing, a bit of reading, a bit of cooking, but nothing I couldn't give up in a hurry if serious work beckoned or I had to move. That might be about to change.

Last week, while wandering the halls in a post-nocturnal-data-taking funk, I saw a sign from the SAA advertising a large-format photography workshop. On a whim I signed up and, after some scheduling and communication glitches got sorted out, spent Sunday learning to work with old-fashioned view cameras with bellows and a black cloth beneath which to stick your head. Also, tangentially, having a look at the SAA darkroom and some of the people who are involved with it.

I'm in deep trouble.

There is a moment of jaw-dropping awe when you point a huge 8x10 camera at a scenic piece of Forest Hills cemetery, stick your head beneath the cloth, and crank the focus until the tree in the middle just jumps out at you with each fluttering leaf picked out in perfect detail. Not to mention the flutters you get in your stomach the first time you take a deep breath, triple-check that the shutter is closed and cocked, pull out the dark slide to expose your precious sheet of film and trip the shutter release. The instructor joked beforehand that he was offering us free crack and now I know why: large-format is a monumentally inconvenient photographic medium, but one look at that ground glass and I was pretty nearly hooked. When I grow up, I want one.

So between that and friendly people dangling all kinds of cool workshops and classes in front of my nose and encouraging me to get involved with the SAA darkroom (I hadn't been in a darkroom this past decade), I'm wondering how to free up some of my schedule for this winter.

I'm putting together a more compact and portable surrogate form of me to participate in a party I'll be very late for. Here's how it starts:

1 cup sugar
1 cup water
2 cc ginger, finely chopped
a moderate shower of cinnamon
1 lemon, zest and rind

Boil

I'm not sure how it'll work as a start for cranberry squares, but it could probably cure your sore throat.

I happen to belong to a flickr group for pictures of Ultra-High Vacuum systems (don't ask). Recently some pictures have been showing up there of a new ion trap being set up in Oxford. Interesting, rather revealing pictures. I wonder if any of my ion-trapping colleagues might be intrigued thereby ...