Brain Scrapings

My laptop almost died yesterday, and I did not mourn its passing. So that means that its time must nearly be up. It's easy not to care when all your data is mirrored across four systems, a couple of back up disks and a flash drive. Some of them are even offsite. Anyway, once the machine had dried out, it seemed to work about as well as it did before, providing I set the date back from 21162. Apparently, a lot of applications crash when it is nineteen thousand years into the future. So that's something to look out for.

Meanwhile, I am no longer accepting things to do for 2007. With one month to go, I no longer have any time available to do anything else in. In other words, if you send me any work to do, I won't get around to it till 2008. If then.

My electricity provider is called "Jersey Central Power and Light". This has been bothering me for ages, and it's only just dawned on me that the reason is that it has the exact same cadence as "Georgie Porgie Pudding and Pie". Presumably this means that I now have to come up with the rest of the rhyme, but I can't think of anything amusing that rhymes with "Light".



Email Woes

Why is it that my email server always dies right at the start of a long weekend? Well, probably the answer is that that when it fails at other times, it gets more rapid attention, or I have other things to worry about and don't notice. Still, it is rather irritating, and makes me wonder if I should look for a slightly more reliable service rather than clinging on to an address from a few years ago.

So instead of finally emptying my inbox, which is currently AWOL, I partook in my annual habit of data mining my email. I must admit, my heart isn't fully in it this year, for some reason, so just a few brief comments.

* Spam is UP! Or else I can count it better. I switched over to using spamassasin about a year ago, and as a consequence, rather than my old heuristic for counting spam (look for senders who sent me only one message), I instead switched to a new system whereby I just count the number of messages trapped by spamassassin. So this is probably a slight underestimate, since I still get a few false negatives but hardly any false positives. My previous estimates were about 1500 per month, but I'm now happily clocking 3000-6000 spams per month.

* Legitimate mail is UP! During 05 and 06, I was processing about 1200-1500 messages a month (and many of them were spam, since I had a lot of false negatives with my old manual procmail spam filtering). This year, I've been clocking 1600-2000 messages a month. Some of these are from heavy mailing lists, but that doesn't explain everything. I must just be more popular all of a sudden.

* Certainly, I'm sending more. One month I sent 300 messages (and this undercounts by loads, because I'm only looking in sent mail, and not in folders which some sent mail is automatically placed in based on recipient). Generally, the trend seems to be up and up. Amusingly, I used to send about 150 messages a month during my undergraduate days; the only exception was Novermber 1995, when I sent 230 -- and that was my second month of having email. Clearly must have had a burst of excitement that month.

* Am still missing loads of data, since I only have procmail logs for my academic email account, not my work email. So goodness knows what the real story is.


$1 for your thoughts

Hmm, apparently, I have to pay an extra dollar to hear amy swearing.

(Back to Black

Back to Black by Amy Winehouse (Audio CD - 2007) - Explicit Lyrics
Buy new: $13.98

Back to Black [Clean Version]

Back to Black [Clean Version] by Amy Winehouse (Audio CD - 2007) - Clean
Buy new: $12.99 )

Things I Don't Like #21491

Sketch shows that have the same sketches every week, and have a little titles sequence before each sketch.


Britain, prepare yourself

OK folks, here are the dates for my upcoming trip to Englandland:

Dec 6: Bleary eyed in LGW
Dec 7: Cambridge
Dec 8: Cambridge
Dec 9: Cambridge
Dec 10: Maidstone
Dec 11: Maidstone
Dec 12: Maidstone
Dec 13: London
Dec 14: Longborough
Dec 15: Longborough
Dec 16: Longborough
Dec 17: Longborough
Dec 18 - Dec 22 Open
Dec 23: Maidstone
Dec 24: Maidstone
Dec 25: Maidstone
Dec 26: Maidstone
Dec 27 - 29 Open
Dec 30: Back to the United States of USA.

If you want to plug any of my gaps, then let me know.


Embed this, mofo!

Oh, I've been really struggling with my fonts. I'm sure I've complained about this before, but I can't find the reference. A number of idiotic publishing companies (ahem, IEEE and ACM) have recently started getting really snotty about embedding fonts into files. In particular, they complain like crazy if you don't embed fonts like symbol and helvetica into your PDF files when you send them to get published. This is particularly idiotic, since these files are part of the "14 base fonts" which every PDF reader must be able to render without needed the fonts to be embedded. So it's a pretty ridiculous rule, and I've wasted some amount of my life whinging at them for having such a rule. In particular, because some perfectly nice PDFs generated by pdflatex have been rejected for this reason. These folks wouldn't give any useful suggestions about how to fix the problem, other than "send us the PS version, and we can generate the compliant PDF." But I'm using PDFLaTeX -- there is no PS version (not even a DVI version). Generating PS would be silly, I'd have to use pdf2ps or similar.
Finally today, I have worked out a clean solution which does not offend me.

Firstly, it turns out that most of the problem is actually coming from the embedded graphics. pdflatex itself does seem to satisfactorily embed most fonts, although you can force it to if necessary, and there are various sites around which say how to do this. But this doesn't help included PDF figures which don't have fonts embedded already. For xfig and so on (another tip: in xfig, the "view --> portrait/landscape" option will fix it so that your figure is output with the correct orientation if it's coming out twisted when you export it), just export direct to PDF and it should work. But some figures are generated in eps format, and then converted to PDF. Usually, I use epstopdf to do this, since it gets the bounding box right, but it doesn't seem to do the trick. So here is what you need to do: issue the command
export GS_OPTIONS=-dPDFSETTINGS=/prepress

and then all will be well. You can work it out for yourself why this will work.


Stating the Obvious

A letter falls into my hands from my gas supplier, with the following message emblazoned on the front: "Forecasts call for below-normal temperatures this winter".

I know what they mean (since they say the same thing every year), but that's not actually what they've said.


Paxman, you idiot

More shocking ignorance from Paxman on University challenge, again leading to him claiming that a valid answer was incorrect. This time, from the program on 22nd October 2007:

Paxman: In statistics, data that are binomially distributed, individual values may be placed in one of two mutually exclusive categories so that the sum of the probabilities of occurring in the categories is what value?
Liverpool Ling: Er... unity.
Paxman: No, it's 1, or 100%
LL: (rolls eyes)

Something should be done!


Near Miss

I was getting very confused, since I had mixed up "Show Me Love" by Robin S with "Show Me Love" by Robyn. Youchoob embeds below: