Planet JCR

I’ve not been sleeping all that well lately. And let’s be honest, this probably has something to do with being in the middle of a term at Oxford University, with a lot on my mind and far too much to see to.

However.

I’ve noticed recently that my room contains a surprising number of light sources, which I suspect can’t be helping me get to sleep. Reading from left to right, we have:

• Light leaking under the door from the staircase (where the lights are apparently permanently on)
• Blue power LED flashing on my laptop
• Green power/charging LED on my other laptop
• Huge red light on the multiplug under my desk
• Flashing lights on my hub

I’ve therefore come up with a cunning plan:

• Block the base of the door with an old pillow
• Prop a book over the laptop light
• Put the other laptop in a drawer
• Tie an old sock round the multiplug
• Shove the hub into a cardboard box with holes for the wires

Whether or not this helps remains to be seen. But here’s a plea to manufacturers: please think twice about the brightness and location of your LEDs, especially if there’s a chance someone will be trying to sleep in the same room as the device they’re on.

I finally got round to putting Linux on my Advent 4211 yesterday. There are various ways and means of doing this; the ones which worked for me were:

• Use the installation image optimised for small screens
• Follow these instructions to make a bootable USB drive from the image (note: it’s ‘cat’, not ‘zcat’ as the ubuntu image is not zipped. I know they’re Debian instructions, but they seem to work)
• Use the script given in this forum post (run as root) to get the WiFi working.

The end results are really rather impressive; here’s a pretty picture:

Ubuntu Linux running on my Advent 4211

Here’s one for all you progammers out there: what happens to the values of x and y if you type the following in, say, Python?


>>> x = 5
>>> y = 6
>>> x,y = y,x


I’d always half-thought, without really considering it, that you’d end up with x=6 and y=6, i.e. that the above was syntactic sugar for:


>>> x = y
>>> y = x


Actually, though, it turns out…


>>> x=5
>>> y=6
>>> x,y=y,x
>>> print x,y
6 5


Do you write Python (not very well?).

Ever wondered why your calls to os.execvp() and its ilk don’t work as expected?

The answer may be that you’re doing this:

os.execvp('ls', ['/home/fred'])

When you actually mean this:

os.execvp('ls', ['ls', '/home/fred'])

As ever, reading the Python docs (help(os.execvp)) is usually better than randomly googling for an answer.

One of the topics that popped up repeatedly at DjangoCon last weekend was how bad purely normalised table structures are, and how denormalisation is good for many things, including making your database cry less.

To that end, during the gaps in PyConUK this weekend, I decided to see how easy it would be to write a new Django field that will automatically denormalise a field in a related table across to another model.

Let's take an example; imagine we have this kind of model:

class Picture(models.model):
    user = models.ForeignKey(User)
    user_username = DenormField(User, "username")
    image = models.ImageField(upload_to="foo")
    title = models.CharField(max_length=100)

The idea is that the user's username (which, let's face it, you're going to need almost every time you need the photo) is stored in the same table, saving you from doing that join to get it from users each time. While this may not seem like much, it can be useful, and besides, it was fun to write.

My DenormField simply listens to a trio of the relevant signals, and then updates this value whenever it detects that the related User object has been updated, so it only costs on write (which, let's face it, is much better than on every read). You can do this kind of copying in normal Django app code as well, but I think this solution is much cleaner. I would do, I guess, since I wrote it.

It's only really a little test exercise in writing Strange New Fields (and I wanted to see how much monkeypatching it would take - none, as it turns out, just some very nasty code tricks). You can get a copy of it if you feel like having a play (warning: only briefly tested; may attach captions to your cat).

There are not very many things that could tempt me to fly 10 hours twice for only a weekend, but DjangoCon was one of them, and I'm glad I went; it was one of the best weekends I've had for quite a while.

There were plenty of good talks; I can't mention them all here, since as we all know, bytes cost money, but in that vein I must at least mention Cal Henderson's talk 'Why I Hate Django' - a brilliantly done talk on the things Django's missing, as well as the fact we don't have a mascot (or do we?) and that we aren't smug enough.

Still, I enjoyed many of the other talks as well (what a shame it is not to be able to see them all in person; still, Google will hopefully release the videos soon), as well as our Schema Evolutions panel, which seems to point towards a bright future for Django and migrations. 

Talking of bright futures, judging by the final session with Adrian and Jacob, there's plenty of stuff to get working on for 1.1, some of which have me quite excited. Multiple database support is going to be one of those important ones, really, but then Malcolm seems to have snuck the bottom layer of it in already, which is quite handy.

All in all, an excellent weekend, and finally a chance to meet a subset of everyone involved in Django (and they really do come from many places), and I can now hardly wait to get over to Prague next year for the promised EuroDjangoCon. If you didn't make it, then I highly recommend watching the videos when Google release them (after editing them on their ridicously great AV system, no doubt).

I bought myself an Advent 4211 last week (you might know it as the MSI Wind; they’re essentially the same thing). So far, it’s lived up to expectations; there are just a couple of things to say…

• Mine seems to have the decent Synaptics touchpad - read The Register’s review for the warning about the other one.
• You really need to use headphones with it if apps you’re listening to don’t have a very good sound level - inevitably in a device this size, the speakers aren’t great, especially at full volume.

How the Linux install on it goes is something I shall let y’all know shortly…

I’ve been aware for some time that my DNS isn’t quite as securely configured as I’d like. http://crashrecovery.org/named/ looks pretty good, but the two main issues bugging me were:

1. Anyone could do a ‘dig @ns.dnorth.net dnorth.net AXFR’ to retrieve a listing of all my DNS records - not great from a security point of view. This is a capability that should only be turned on for secondary DNS servers which need to fetch from the master.
2. The server would perform arbitrary lookups [for any domain] on request. This means it’s operating in ‘recursive mode’, which is a Bad Thing for various reasons.

The solutions were:

1. Add “allow-transfer { “slaves”; };” (without the double quotes) to the section of the configuration beginning “zone ‘dnorth.net’”. Then add a section defining the “slaves” access control list to be the local server, plus the secondaries: “acl slaves { 127.0.0.1; 123.45.67.89; }” replacing 123.45.67.89 by the IP address(es) of your secondary nameserver(s).
2. Add “recursion: no;” to the “options” section of the configuration.

Then restart the BIND9 service - on Debian, this is “/etc/init.d/bind9 restart”.

Health warning: Don’t do (2) above if you rely on your server to do its own DNS resolution - follow the crashrecovery tutorial above instead.

I did a lightning talk to CompSoc last night on whether 64bit is ready for use on the desktop - if you’re bored, you can download the slides and the second of my famous sketch and scanned graphs.

gq has been a necessary evil in my life for some time now. I need a graphical LDAP client for use on the CompSoc systems, but gq (to be fair, the versions of gq packaged for Ubuntu) seems to be very buggy, segfaulting all over the place if you try to do anything other than browse with it.

Last week, after upgrading to a 64bit version of Ubuntu for the first time, I finally ditched gq, after running into identical symptoms to this Debian bug.

The good news is that there’s an alternative that actually works: it’s called Luma.

Have fun.

The epic tale of how my new mail setup was born

My personal email has been on a rough ride over the years: from a reasonably nice (Microsoft based!) school email setup in 1999, which sadly got removed when Windows 98 was introduced, I went through three Hotmail accounts. I had a brief flirtation with GMail, but not being all that keen on the means of delivering advertising, I ended up back on Hotmail.

Registering dnorth.net last year at least ensured my email address would no longer change, but the technical capabilities of the two mail servers holding the mail still left much to be desired: flaky, unreliable spam scoring, no facilities for server-side filtering/sorting, sheer lack of customisability…

Last week, I finally did something about it. At my disposal was my VPS, running Debian Linux 4 (’etch’). On the wishlist were:

• Accurate server-side spam scoring with SMTP-time rejection of the most obvious spam
• Sender verification
• Sieve filters for server-side sorting into folders
• All mail stored on the server and accessible over secure IMAP

Thankfully, none of the above is too difficult: some pretty good instructions are Out There for most of it. The ones I used were:

• Exim4 on Debian with SpamAssassin, ClamAV, Virtual Domain Alias Files and Message Size Limits per Domain
• Exim + Sieve + Avelsieve Howto
• Handling mail for multiple virtual domains with exim4

Please remember, I am not responsible for the content of external sites (e.g., the links above), nor can I accept any responsibility for the consequences of acting on the points below…

I ran into a couple of issues:

• Permissions on the .sievesource files generated by AvelSieve - I needed to chmod g+w on /var/lib/squirrelmail/data and chown it to www-data:www-data in order to reach a state where Exim could read the file, and Avelsieve could write it.
• Exim4’s native sieve implementation only has the core features in it, not the extensions defined in RFC 3431. I dodged the issue by matching the number of *s in the X-Spam-Score header using string matching, rather than numeric checks on the X-Spam-Score.

All in all, though, it’s working a treat. Email perfection at last!

Twelve months ago, I installed Linux (Kubuntu) on my laptop.

Twelve minutes ago, I got the wireless card working with my WPA-PSK home network.

A combination of the progress made on drivers over the last year, and finding which of the squillions of instruction sets out there Actually Worked has finally got things going - rather than recount it all here, I’ll point owners of the HP Pavilion dv5157eu to this page, and advise everyone else as to how I found it: do a Google search for some combination of ‘ubuntu’, ‘wireless’ and your laptop’s exact model number.

Edit: My wireless happiness sadly didn’t survive a hibernate or a reboot. Turns out the line of the page which says do ’sudo echo ndiswrapper > /etc/module’ actually means, as far as I can tell, that you should do:

sudo -s
echo ndiswrapper >> /etc/modules

As ever, follow my advice at your own risk.

Musings on public transport, and a look round The Other University

So, full term ended yesterday, and I had to get over to Cambridge for a dinner. The X5 bus proved reasonable, both in terms of cost (£15 open return) and time, though the long stop to change drivers in Bedford lengthened an already long enough journey.

After a very pleasant afternoon being shown round Cambridge, I’ll have to begrudgingly admit to being impressed by it. King’s College Chapel, in particular, was spectacular. Even the weather wasn’t bad.

The train journey home from Oxford was tolerable, though prolonged by a needless half hour stop in Birmingham - it wasn’t so much the stop that annoyed me, but the fact that nobody bothered to tell us about it.

Yesterday only served to reaffirm my belief that public transport is mediocre at best, and ghastly at worst. Our rulers will have to try harder if they want to convince me not to use the driving licence lurking in my wallet, as soon as I can afford to…

A problem which has been bugging me for a while is my Konsole and Yakuake terminal sessions ceasing to respond to input for no apparent reason. The symptoms are always the same: the session seemingly stops responding to key presses, yet if I was running screen inside the session, reattaching it elsewhere reveals the characters from the missing key presses.

The answer turns out to be that hitting scroll lock locks the terminal so that the arrow keys scroll it rather than cycling through previous commands, and this also prevents key presses updating the display. My Yakuake hotkey is F12, which is next to scroll lock on my laptop.

So now you know.