The Bucket O' Nothing

My post this year is about someone who I'd not heard of until I started my current job, but whose work was a vital part of the computers I used for most of my school years - the BBC Micro, which we used at junior school, and the Acorn Archimedes, which we used at senior school for the first four years I was there.

Sophie Wilson was not only one of the hardware engineers who designed the BBC Micro and its operating system, she designed the BBC Basic dialect for it and wrote the interpreter .

She designed the instruction set for the first ARM processors, the ones that went into the Acorn Archimedes. I've spent some time looking at the instruction set as it was in ARMv7, and comparing it with the instruction set as it was in the Archimedes it's impressive how similar things are.

(Side note 1: The patents on the version of the instruction set architecture from back then (ARMv2, I believe) have now expired, and so there is an open-source ARMv2-compatible processor project that uses the design from then. It is quite cool.)

(Side note 2: ARMv7 isn't the latest ARM architecture - that's ARMv8, which complicates things by having an extra new ISA. It does still have the original ARM ISA, though. With the extra things that have been added since ARMv2, of course, such as a divide operation, vector instructions, some mnemonic renaming, the entire Thumb (now T32) instruction set, and I could go on. But the basic structure of what's now called A32 - the 32-bit ARM instruction set - has stayed the same in several ways.)

Since Acorn/ARM, she's continued to work in processor design, including designing the FirePath processor that's used for digital signal processing for DSL.

(Side note 3: I found a presentation on FirePath that included the phrase "Viterbi acceleration" which makes me unreasonably excited because oh wow, I remember the Viterbi algorithm from doing statistical language modelling in my MPhil and have only now realised that of course it has more general applications for finding highest probabilities than just state transitions through possible phonemes. Yay, putting disparate bits of knowledge together. :D)

For all the work she's done in processor design, she's been appointed as a Distinguished Engineer at Broadcom (where she currently continues to work on Firepath), a 2012 Fellow of the Computer History Museum in Silicon Valley, and was a finalist in the 2013 European Inventor Award.

EDIT: Correcting something I said initially - winterlove pointed out that she wasn't one of the founding members of ARM the company itself, as she stayed with Acorn when ARM split off. (Wikipedia is the only place I've managed to find an actual list of the founding members - most articles just give a number and no names.)

Read The Act of Roger Murgatroyd by Gilbert Adair while in the pub today (about 50 pages while waiting for people, and the remaining 200-ish while eating and chatting).

Link above goes to the Guardian review at the time, which sums up my thoughts. (And contains spoilers.) The dialogue is clearly a weird blend of modern slang and attempts at sounding 1930s-ish, and while this might fool someone who's never *read* any Christie novels, to me it sounds like bad fanfiction both in terms of language and what topics are seen as acceptable to discuss (and euphemisms that get used instead). The fourth wall breaking is clunky, the repeated references to John Dickson Carr do not impress me as they are evidently intended to (and actually, no, the chapter about locked rooms in The Hollow Man wasn't delivered by the author, it was delivered by the detective and did manage to make sense in character).

The reveal of the murderer made me go "oh for fuck's sake" but I can at least respect what the author did - to the point that I was considering going back and rereading the last 240 pages to see if he'd done what I thought he had - and then one of the characters broke the fourth wall again to explain it.

On the other hand, I did think the armchair idea was quite neat.

---

I'd recommend The Murder of Roger Ackroyd by Agatha Christie instead. The characters are more entertaining and consistent, the dialogue isn't straining to be "period" (it is instead appropriate for its setting), and there's no crappy fourth-wall breaking.

Also, the anime series Hyouka, featuring someone who's actually quite good at solving mysteries but is also resolved not to actually expend any effort on school. Of course he ends up solving mysteries anyway. :)

Empire continues to be fun as I discover how much stuff we can get out of going up to people and saying "hello please tell me about your character('s past life)" and looking interested. Also I think I have found more comfortable kit, if slightly more OOC-looking right now. Also also I only got a little burned and my tent did not leak even during the rainstorm, and my kit mostly dried out from that just fine.

This afternoon loopymushroom and I took photos of our kit in her back garden and back room - they're on Flickr here.

Kit search for Empire took me three trips to charity shops: one with freddiefraggles to Burleigh Street, one with loopymushroom to Burleigh Street, and one on my own that went along Burleigh Street and all the way up and down Mill Road. (There are so many charity shops in Cambridge!)

Other things are a dress that's been in my wardrobe for years (wore it as a teenager), leggings from Sainsbury's, and an excellent embroidered abaya/jilbab that I bought on eBay.

( Photos under the cut, and an awful lot of notes about costume. )

Python Software Foundation moves to require that Python conferences have a Code of Conduct if they want grant money from the PSF for running the conference.

(In this case a code of conduct is a written statement, usually on the conference website, saying what behaviour is and is not okay at the conference. The PyCon US one is pretty good and has been borrowed or adapted for a lot of other Python conferences. It contains a useful but not restrictive definition of things that constitute harrassment, a list of people to talk to about harrassment or security concerns, and statements of what the conference will do about it.)

A PyCon IE attendee feels a little down about this, as it seems to be too "formal".

The first couple of paragraphs saddened me (I was worried it would be a whole post of "I've never encountered this problem, so I don't think it exists"), but I'm kind of glad for the last few:

I don’t know, makes me think about what is the general perception and behaviour of the development community. I know there is discussion out there about whether the geek population is welcoming to diversity or just a bunch of jerks that just can’t behave (and all the spectrum in between). I guess it just makes me sad to think that we may need "an adult" telling us not to say things that we already know that we shouldn’t. It’s 2012, we have no excuse.

As I say, I just feel a little... disappointed. Like thinking that there is something wrong in all that, that we are grow up and that things are not on the same level of friendly informality. That we need rules to ensure everyone feels safe. I guess that a small number of spoiler brats are just ruining the party to everyone else.

And Jacob Kaplan-Moss's comment on the post is excellent (he's written more about Codes of Conduct on his blog):

I think the discomfort you’re feeling is totally legit and valid – codes of conduct are kinda patronizing to the vast majority of people who know better. The fact that you don’t need a CoC to tell you how to behave is a good sign: it means you’re a Normal Person who understands how to make other people feel comfortable in your presence. But you have to understand that the CoC isn’t about you. It’s about making other people comfortable in settings that have traditionally been quite hostile to them.

---

As for me, I'm happy about this change - it means that maybe local-to-me Python conferences might not see Codes of Conduct as an afterthought or something nice but low priority. I'm not even sure whether PyCon UK got its Code of Conduct written before the "conference song" someone wrote.

The story so far. Not as bad as the Ikea or Virgin shenanigans I've heard from Aquarion or Dave, but bizarre in its own way.

1. Ordered a few things from ASOS last week. Delivery company was DPD, who give you a time window for when they'll deliver and can reschedule it for you.

2. Scheduled delivery for Saturday at work address, because I've had things delivered on Saturday before so I assume there's usually someone on front desk or post duty to sign for it. Saturday came, I got an email saying nobody collected it so I should sort out redelivery.

3. Went to the DPD site to reschedule, with the card number they'd given me in email. Discovered that the package had apparently been signed for - in February.

4. Phoned DPD, where the guy I talked to said that the card number had been recycled and (through bad luck, I guess) had previously been used for a delivery to the same postcode (i.e. my office). He booked the redelivery for Monday, and mentioned there were some other parcels heading to the same address.

5. Monday came with a text message telling me the parcels would be delivered around noon, and Monday went with nothing from DPD about them having delivered the parcels, and nothing from our facilities department about me having a parcel.

6. Tuesday came with a text message telling me the parcels would be delivered around noon, with a slightly different hour-long window but the same parcel number as yesterday.

7. Then someone at work emailed me because he'd got a reminder email from DPD about a delivery to the office in my name.

We've exchanged a few emails, and he's waiting for a delivery that wasn't collected on Saturday as well (although from a different online shop - to be extra confusing, he has also this morning ordered from the one I ordered from last week).

I think what DPD might be doing here is indexing orders by delivery address minus name, which strikes me as the stupidest damn thing if you stop and think about the existence of offices. Maybe adding a date, which still doesn't solve anything.

If the parcels don't turn up this lunchtime, I'll phone DPD and see if I can find out what on earth happened.

It appears to be mosquito season and the bastards keep biting my arms while I sleep. Is antihistamine cream supposed to do something other than make the bites swell up massively?

(Thankfully I don't think any of the bites will get in the way of lifting; they're on the backs of my forearms and an elbow this time round.)

Read "If On A Winter's Night A Traveller" a few months ago - it started out really interesting and then hit several of my no-stop-wait-goddamn-casual-thoughtless-anti-feminist-bollocks buttons and I ended up swearing at it once I put it down.

(I had a rant about it, I'll put it up at some point, either here or on a newer "here is where I am putting all the feminism stuff" blog.)

Was planning to read it through again to see if I had just imagined the main character being male, or if it would read fine with them being female-and-interested-in-women.

Turns out I didn't imagine it, they're identified as male on page 32. :( Ah well, exercise in creative rereading squished.

(The pages before that were kind of weird when read with that in mind; most of them were fine, and then the gender stereotypes demonstrated in the text but not referring to the main character would collide with things that the main character was suggested as possibly doing, such as the job they worked at.)

I have a tiny website working. It is a ridiculous thing that combines random panels of a terrible webcomic.

It's the first thing I've ever hosted on a server I administrate myself, so here are some notes mainly for my reference.

Setup was Ubuntu 10.04 / Apache / mod_wsgi / Flask, with the Python Image Library to do the image extracting and combining.

Basic docs
* Linode Library has a basic guide to Apache configuration and a guide to setting up mod_wsgi on Ubuntu.
* Flask's documentation also has a guide to deploying with mod_wsgi; like the example above it uses virtualhosts, but the suggested file is different.

I ended up using a combination of the two.

Steps needed (very basic, fragmented, will refine)
* Make a directory in /srv/www/ appropriate to your site or something.
* Stick your files in there. Probably have your site stuff in a subdirectory.
* Put an apache.conf file in there; crib most of it from the guides above. Things you need to put in there:
- your domain name (ServerName)
- what path under that domain name the WSGI handler is going to use and what .wsgi file it's going to use to read it (WSGIScriptAlias)
- a DocumentRoot and Alias entries if you want Apache to serve some files directly rather than passing them through WSGI, e.g. robots.txt and static files like CSS
- ErrorLog / CustomLog entries for logging (see below)
* Make a some-appropriate-name.wsgi file in there too. Crib its contents from the guides above. What it needs is:
- your Python site's directory in sys.path()
- the Flask app object in your main file imported as "application"
* You now have stuff for a VirtualHost almost done but Apache knows bugger-all about it because it's in some random directory. So, two steps:
- ln -s /srv/www/yoursitedirectoryname/apache.conf /etc/apache2/sites-available/yoursitenameorsomething
- and then a2ensite yoursitenameorsomething - it uses the name you put for the simlink here, so make sure you haven't typoed it if you get a "Error: yoursitenameorsomething does not exist!". (It turns out spelling "bouteillebl.eu" without leaving out the second 'e' is harder than you think.)

Error logs
* I have an ErrorLog going to error.log and a CustomLog going to access.log in my apache.conf for this site. There are also the /var/log/apache2/ logs.
* I also stuck LogLevel info into my apache.conf so there's a slightly greater amount of detail in my site's logs.
* From my attempts to get stuff to work, any errors that are thrown at "compile time" (as in, syntax or import errors that are thrown when the .wsgi script attempts to import my Flask app) turn up in my site's error.log. Errors thrown at "run time" (as in, when an HTTP request comes in but something in my code throws an error or exception) end up in Apache's logs.
* Actually they don't even do that unless I add

import logging,sys
logging.basicConfig(stream=sys.stderr)

into the .wsgi script, as otherwise the errors get swallowed up.

Actually getting those errors logged was the biggest problem, as before I managed that I didn't have anything to go on.

It turned out that (a) I was pointing at the wrong part of the filesystem when looking for the image files and (b) the directory for the cached image files wasn't writeable by Apache's default user. Both of these were pretty easy to sort out once I knew what I was doing.

Heads up for anyone on Vodafone pay-as-you-go with an iPhone, or who's thinking of going for this. This may be a similar issue for anyone with a smartphone (at least, the first point).

a) Vodafone's PAYG mobile internet settings are not the same as their pay-monthly ones (as noted in this post on the Vodafone forums, the APN is pp.vodafone.co.uk, not wap.vodafone.co.uk as for pay-monthly).

b) If you buy an iPhone from an Apple store it will be set up to use the pay-monthly settings because seriously, who buys an iPhone and then spends less than £5 a month on calls and texts (having run the numbers on the call logs from their old phone), of course you'll have a contract. If your new phone doesn't work despite a Vodafone rep telling you it's been set up to have internet access, this may be the problem.

c) I installed the iOS update over wireless at the weekend, and somehow the mobile internet settings got set back to the monthly ones; if you've suddenly lost internet despite your phone telling you it can find a 3G connection, this could be the issue.

Quick reference:
* for pay-monthly the web and MMS APN is wap.vodafone.co.uk and the user/pass is wap/wap
* for PAYG the web and MMS APN is pp.vodafone.co.uk and the user/pass is web/web

The rest of the web/MMS settings are the same.