Beware of the Train

Thanks to all who helped out with my ssh problem the other day. A few different approaches worked: I'm writing them up here for easy reference.

To recap, the problem was as follows: I want to ssh from my home machine (delight) to my work machine (sentinel) without typing any passwords (and, while I'm at it, to various other work machines, such as the Subversion host). Unfortunately, sentinel isn't visible on the public Internet; first I need to ssh into a gateway machine (rydell) from which sentinel is visible. Oh, and neither sentinel nor rydell allow public-key ssh login; both require you to use the Kerberos authentication protocol, which is explained here.

Brute force and ignorance

My first attempt was to use ssh rydell ssh sentinel (which sshes into rydell, and invokes the command ssh sentinel thereon). This failed with the error "Pseudo-terminal will not be allocated because stdin is not a terminal". Marco Fontani pointed out that the -t switch to ssh allocates a pseudo-terminal, so ssh -t rydell ssh sentinel Does What I Want.

X11 Forwarding

Mat Brown pointed out that I was overthinking it: by using the -X and -C options to ssh (or adding the lines ForwardX11 yes and Compression yes to the relevant stanza of ~/.ssh/config) I could enable compressed forwarding of X11 connections; I could then create new terminal windows by sshing into sentinel once and creating new xterms on there. I already had ForwardX11 set, but didn't know about Compression, so I've enabled that; it seems to help.

Using ProxyCommand and AutoSSH

I added the lines

ControlMaster auto
ControlPath /tmp/ssh_mux_%h_%p_%r
ServerAliveInterval 60
ServerAliveCountMax 60
Host rydell
    User pvlak1
    HostName rydell.my.employ.er
Host sentinel
    User pvlak1
    ProxyCommand=ssh rydell nohup nc sentinel 22
    HostName sentinel.my.employ.er
Host svn.my.employ.er
    User pvlak1
    ProxyCommand=ssh rydell nohup nc svn 22

to ~/.ssh/config. Then, at the beginning of the day, I set things up with the commands

kinit pvlak1
autossh -f -M 0 -N sentinel
autossh -f -M 0 -N svn.my.employ.er

I can now open a new ssh connection to sentinel in an eyeblink. I needed to install AutoSSH, but this was just an apt-get away.

Two things are going on here. The ProxyCommand lines (suggested by hatfinch, and debugged by simont) tell ssh how to reach sentinel: in this case, by sshing to rydell and using netcat to open a connection to port 22 on sentinel (on which sentinel's sshd is listening). The HostName line is necessary to stop Kerberos getting confused. The first four lines were suggested by Marco Fontani and Aaron Crane, and allow ssh to multiplex all its connections to sentinel (or any host, come to that) over one channel, eliminating the need for a cryptographic handshake on each new connection and leading to blazingly-fast startup times. To avoid various annoying problems with this setup, you'll need the AutoSSH invocations: Aaron explains why on his blog.

My original problem is solved - hurrah! Now, can anyone explain why resize events weren't being passed through my expect script, and what I could have done about it? :-)

According to the book Invisible on Everest: Innovation and the Gear Makers, the noted fell-walker and guidebook author Alfred Wainwright had four suits: his best suit, which he wore to church; his second-best suit, which he wore to work; his third-best suit, which he wore for hillwalking; and his scruffiest suit, which he wore for gardening. When his gardening suit wore out, he'd buy a new suit, which would become his Sunday suit, and all his existing suits would be demoted to the next position in line. This was in the days when business attire was still constructed sturdily enough to stand outdoor use, and specialist mountain clothing (made of silk and gabardine) was outside almost everyone's budget. I suspect that Wainwright's approach to clothes-shopping was fairly common at the time, though Wainwright's suit-pipeline was probably longer than most.

---

Many years ago, probably some time in the mid-Nineties, I watched a ten-minute documentary on (I think) Channel 4, about two craftsmen who manufactured their own knives for use in their work. One was an up-and-coming sushi chef, and the knife he was making was his first; I think it may have marked the end of his apprenticeship, but I can't remember much about his segments of the programme. The other was a maker of ballet shoes, with many years' experience. In the course of his work, he used half-a-dozen different types of knife, each with its own specific use. But the thing was, they were all the same knife. He made his knives on the Wainwright principle, you see. His knives all had to be razor-sharp all the time, so he made them from very soft steel, and sharpened them constantly. Cut, cut, sharpen, cut, cut, sharpen, switch to new knife, cut, cut, sharpen, etc. The blades thus wore away very fast, so what had started out as a large convex blade soon turned into a smaller straight blade, which then turned into a thinner concave blade, and then the tip snapped off and it became a different type of blade again. Roughly once a month, his smallest knife became useless or his largest knife stopped being useful for the largest-knife jobs (I can't remember which), and he knew it was time to forge a new large convex knife.

It was a truly great piece of television, opening doors into little worlds I'd never imagined. But I can't find it now. Living as we do in the far-off future of the twenty-first century, I'd have expected someone to have uploaded it to the 'net, or at least written about it somewhere, but I can't find it at all. Worse yet, it was part of a series; I have no idea what the other episodes covered (I think the overall theme was something like "how people in unusual jobs do their work", or maybe "the tools people use"), but I'd love to see them.

Can anyone help?

We have mice (or mouse, at least). We'd rather not. Hints for getting rid of them, anyone?

A while ago, I was playing about with my flatmate's shiny new eeePC. Impressive, I thought, but the keyboard's a bit of a fiddle - the keys are just a bit too small for my fingers (which are neither fat nor stumpy). Plus, the machine seems practically custom-designed to be balanced on one hand and operated with the other, and typing on a proper keyboard with only one hand is nothing short of an offence against all that is Good and Righteous. Surely there's a better method for one-handed text input onto such a small device?

Well, yes there is, as it happens. I'm talking about chording keyboards. In a normal keyboard, each key stands for one character, and your fingers move about from key to key; in a chording keyboard, your fingers stay fixed, and you indicate different characters by pressing combinations of keys. It's a generalised version of what the shift key does. Chording keyboards apparently take some getting used to, but allow for very high-speed and natural text input. Here's Stephen Fry talking about the chording keyboard on an early PDA:

My obsession with SmartPhones began many, many years ago. Certainly well before such devices existed in the real world. From the first Sharp contact-and-calendar “electronic organisers” , through the early Psions, the sadly missed AgendA (see above: no QWERTY there, more a kind of weird courtroom stenographer’s chord-based input pad: never have I been able to write faster than with that splendid object - I had another device using the same input system called, I think, Qinky, which connected to the Centronix port of a BBC Micro), to the opening salvo of Palm Pilots, Apple’s Newton and the arrival of Handspring. If they existed I had to have them. Had to.

[The rest of the article's well worth reading. And I see Stephen Fry now has an eee, too...]

Now, hardware chording keyboards still exist, and some of them speak USB: there's a problem, however, in that all the ones I can find are ridiculously expensive. Like, several hundred pounds. There are instructions for making your own online, but my soldering iron's in Oxford and I'd clearly never get round to it anyway. But then it struck me - we've got a perfectly good set of buttons sitting there on the machine already, we don't need any extra hardware! Have the keys H, J, K, L and Space work as a chording keyboard - that's (2^5 - 1) = 31, combinations, which is enough for the English alphabet plus, say, ",", ".", space, "?", and "'". We could deal with capital letters in the way that text-messaging software does, by making the combination ". " put the next character into upper-case, and possibly by making Shift sticky. We can just have the remaining keys self-insert as normal, so we don't have to do anything very cunning about characters that aren't in the subset represented by chords - I probably type the { key a lot more than most people, but I don't think I'd mind switching to one-fingered typing for that. I wasn't really thinking of this as an input method for heavy coding, anyway. The obvious encoding system to use would be Baudot, or something based on it, but I'm open to suggestions.

All fine in theory; but how would I go about writing such a thing? Or does one exist already? I had a brief look, but given all the different meanings of "chord", I didn't see anything very promising. I'd thought about writing an Emacs minor mode to handle the chording, but (a) I don't know my way round the Emacs codebase, (b) I don't actually use Emacs (and does it even run on the eee?), and (c) it seems like the Right Thing would be to use XIM (X Input Methods). Unfortunately, XIM documentation appears to be... well, have a look for yourself. And the thought of writing low-level X code in C fills me with fear and loathing - OK, OK, I'm a wuss who's been spoiled by high-level languages.

Does anyone have any advice? And does anyone fancy doing it for me joining in?

After a thoroughly stressed-out weekend, I finally got the job application finished yesterday at 4pm. I was working on the basis that the recipients were four hours behind me :-) There's probably some vital bit of paper I'm missing - I'm worried about my transcript in particular, which didn't get sent off until last Thursday - but what the hell, it's done now.

Anyway, here's the research proposal I sent, and the one-page summary for layfolk. Thanks very much to those of you who commented on earlier drafts, and in particular to jonjonc, for telling me about Barry Jay's work. I'm not yet sure how relevant his work is - I emailed him, and he takes the usual CS line that if it's not statically typed, it doesn't count :-) He's since moved on to languages with first-class patterns*, which sounds pretty cool.

Think of Haskell's pattern-matching in functions (or Lisp's destructuring-bind). Like that, only you can pass patterns around as variables.

markdominus, who is a mathematician by training and a programmer by profession, has been attending undergraduate physics lectures and coming out more confused than he went in. It's a feeling I remember well from the last time I studied physics, and actually part of the reason I did my degree in mathematics - clearly I was never going to get to the bottom of this stuff if I were taught it by the physicists! :-) But then it turned out that physical applied maths was Really Really Hard, so I retreated back into pure maths.

Anyway, in one of his latest posts, he asks a series of questions about electromagnetism, and I realise to my shame that I don't know the answers to any of them. I did do a course with "electromagnetism" in the title once, but the other half of the title was "relativity", and wouldn't you know it, but relativity took up nearly all the teaching time. Anyway, I'd greatly like to know the answers to these: perhaps one of you physics or engineering types can help?

On the theory that I'll probably gain more understanding by thinking about it myself rather than just asking others, here are ( my guesses. )

So I was down in Cambridge last week, as part of an exercise in convincing mathematics Part 3 students that there are other universities in the UK, and that some of them are even worth doing PhDs at. I didn't manage to see everyone I'd have liked to, but I did get to see Antarctic Mike for the first time in nearly two years (as the name suggests, he spent most of that time Down South), and to spend a night on his new houseboat. I also went out to the pub with Mike and Ros (our musical director from The Matrix) and had a very nice lunch with scribeofnisaba, whence the rest of this entry.

scribeofnisaba is an Assyriologist (coolest job title ever, no?), which means that she studies the literature and languages of the ancient Near East - the languages spoken by the Assyrians, Sumerians, Babylonians and so on. Roughly speaking, the languages spoken in modern-day Iran and Iraq, in the period 4000-1000BC(ish). In other words, the study of the oldest literature in the world. She tells me that it's a very exciting field, and that now is a very exciting time to be working in it, as so little is known, and much of what we thought we knew has recently been shown to be wrong - for instance, it was always thought that Sumerian scribes were all male, but this is now thought increasingly suspect. One surprisingly basic thing we don't know (if I'm understanding her correctly) is whether or not Sumerian poetry was metric.

A brief word of explanation. Poetry in many languages is characterised by the use of a repetitive pattern of long and short syllables, called the "metre". Shakespeare, for instance, used the pattern

SLSLSLSLSL
SLSLSLSLSL
...

whereas Virgil used

LSSLSSLSSLSSLSSLL
LSSLSSLSSLSSLSSLL
...

Now, it seems to me that it might well be the kind of problem that reacts well to having a few billion processor cycles thrown at it, but there are a couple of things I'm not too clear on, and maybe you lot can help me out.

As I understand it, we don't know
0. whether the Sumerians wrote metric poetry at all,
1. if they did, what metre they used,
2. how individual Sumerian words were stressed.

Assuming 0, then if we know 1, we can work out 2, and vice-versa. But if we don't know either, we might be able to bootstrap a knowledge of both by considering their interaction. This could well turn into a lifetime's work if done by hand, but fortunately we now have wonderful machines for doing repetitive calculations very fast, and the Electronic Text Corpus of Sumerian Literature seems like it was designed for precisely this kind of automated search.

[scribeofnisaba: You mentioned that cuneiform was syllabic: is the same symbol typically used in many different words? Because if so, we might be able to assume that it's stressed the same way whenever it occurs. But then again, probably not. Also, do we know that the lines are metrical units?]

Anyway, here's my idea. We search the space of possible ways of stressing words, and looking at the metres they produce on the corpus of poetry, and search for the stress patterns that produce the most regular metres. The searching's not that difficult, or rather there are well-known ways of doing it (more on that in a minute). The tricky bit, actually, is finding an automated way of looking for "poetic" stress patterns - the problem is that the pattern "short short short short short..." is as regular as you like. We could (for instance) calculate the stress pattern of every poem in the corpus, given our current guess at word stresses, then compress the resulting bitstream and observe the compression ratio - but that would tend to produce "short short short short...". We could, er, apply a discrete Fourier transform to split it into a sum of periodic functions, and then, er, um. Just eyeballing it won't work - we'd probably have to go through several thousand (mutate, calculate metre, compare) cycles. Is there some kind of constraint on word stress patterns for Sumerian that we know about? Each word must have at least one long syllable, or something? Anyway, there's presumably some way of doing this that I'm either too stupid or ignorant to find right now :-)

As for the algorithm for the search: the most obvious thing to do would be to simply try every pattern, but with nearly a thousand nouns in the glossary (and many other words of other sorts), each containing several syllables which can be either short or long, we'd have a search space with around 2^2000 elements. This is, to say the least, computationally infeasible. Another approach might be via "hill-climbing": start somewhere random, change it a bit, see if the change improves matters, keep it if so, repeat until you can no longer improve your results by making small changes. The problem with this is that you can get stuck at a "local maximum", or a point better than everything nearby but still not the best overall. Every mountain-top is a local maximum for the height of the Earth above sea level, but not every mountain is Everest. One way of dealing with this problem is to introduce a bit more randomness: the following elegant algorithm (which is the one I was thinking of using) is due to Metropolis:

1. Start at a random point in the search space.
2. Make a small change to your position.
3. See if this has improved things. If it has, keep your change. If not, toss a (weighted) coin, and keep the change if it comes up heads. The coin should be weighted proportionally to how much worse the change has made your results: a very deleterious change has a very low chance of being kept.
4. Repeat several hundred or thousand times until your position stabilises.

I attended a series of lectures by the mathematician and magician (mathemagician?) Persi Diaconis on (among other things) this algorithm about a year ago, and it's really surprisingly effective.

( Cut for geekiness )

So, can anyone help me with my Haskell problem? Anyone? Bueller?

I need puns about the life and work of Robert Burns and maths, and ideally dynamical systems. I feel it would be poor form to steal all my jokes from the Klein Four Group or Tom Lehrer*. It's a tall order, but I know you guys can do it - come on, ExoBrain, do your stuff!

Which is to say...

Weejies: we're having a Burns supper tomorrow evening, from around eight. You're invited. Let me know if you're coming and preferences for veggie or real haggis.

* I think of great Lobachevsky and I get idea, ha ha!

steerpikelet can use your help! She's writing her special topic paper on the internet and the like in fiction (and especially science fiction, AFAICT), and needs booklisting. She describes it thusly:

Under the general bracket of 'fiction in English,' I'm doing an extended essay all about how 20th/21st century literature uses t'internet as a narrative hook, and I'm going to go on a bit about narratology and post-structuralism and probably end up talking about fanfiction. I intend to use the words 'semiotic' and 'schema' a great deal.

Knowing that a lot of you a) know loads about sf, b) don't read her journal, I thought I'd post the message here - her original cry for help can be found here. Thanks in advance!

When I was little, I always assumed that science fiction would be the default reading matter of the intelligentsia - after all, it's SCIENCE fiction. I had a bit of a shock when I went away to Big School and found that it was not only much less popular than I'd thought, it was actually looked down on by many intelligent-seeming people. It's good to have so many intelligent sf readers on my friends list, as it suggests my earlier belief might not be so much wrong as twenty years too early :-)