Back in late 2009, we started writing something we nicknamed “parlytags” – a first attempt at geotagging Parliamentary data, an experiment into what was possible using only existing public-facing data. It was retired today with a view to replacing it with a newer, leaner, faster one (but that’s a different post entirely).
It presented itself as a search box that accepted a UK place name
The app searched a local copy of data from the GeoNames.org project, found the right grid references and rendered an embedded Google Map, along with any suitably tagged data we’d scraped up earlier. The buttons underneath the text were clickable links to other places the data had been tagged with. Partly as a functionality demo and partly to compensate for the small data sample we were using, tagged data for up to a 40km radius was included in the results rather than just direct hits on the name
One of the nice side effects of using the geonames.org data was the ability to match against the alternate place names allowing for non-English names, regional variations, colloquialisms and common alternative spellings to become valid search terms for little extra effort
Where the initial place search results were ambiguous (i.e. the name could refer to 2 or more places), it chose one to focus on initially – the first one on the list, prioritised by the place type if my memory serves – and offered links to the others
The service may be gone now but the code can still be found on Github
December 15, 2008 – 1:26 pm
I changed the blog theme again. Our last theme, nice though it was, didn’t have bylines on the posts. And we really want bylines.
Apologies especially to Will Howells, who was not only the very first commenter we had, but actually liked the previous theme. Sorry Will.
December 10, 2008 – 12:39 pm
To be able to “stress test” our server, we’re preparing to deploy an Open Source application called Siege which simulates multiple users making HTTP requests to the web application. It didn’t do quite what we wanted in the logging department, so we’ve hacked it about a bit and, in the spirit of Open Source, made it available on GitHub.
Now we can get down to writing some test scripts that use Siege to test specific things on the server to see how it (and by extension our application) will behave under load. To make these artificial tests more realistic, Siege helpfully includes the ingenious -i command line option which will make it choose from the list of urls we give it in a cheerfully haphazard fashion, turning it into the server testing equivalent of a roomful of monkeys with internet browsers that will only connect to our site.
(The full command we’re looking to use is: siege -c10 -r50 –file=urls-to-test.txt -E session.log -i in case anyone was really interested in the geeky details)
Now to solve the problem where my plastic laptop starts to melt before the server looks even vaguely stressed…
December 10, 2008 – 12:22 pm
I changed the theme of our blog. Our spartan ‘design’ was hurting my eyes. I’m trying out the ‘Day Dream’ theme – let me know what you think.
November 19, 2008 – 3:13 pm
We attach Dublin Core metadata to our pages: view source on the Hansard prototype – it’s there at the top.
Is this metadata useful? Are we doing it right? Could we do it better?
November 19, 2008 – 1:03 pm
We currently style tables in a slightly smaller font size than the main text on the Hansard prototype site. We’d really like to know if this is working.
Have you spotted a table that was incorrectly formatted? Do you find them readable and accessible? We’d be grateful if you could take the time to make us aware of any odd-looking tables using the Google issue reporting form.
November 13, 2008 – 1:29 pm
We’re no longer recognising geonames in the Hansard text, nor are we producing KML files from sitting days.
Although it was an interesting experiment, it wasn’t useful enough in real world usage to enough people. Removing these experimental features means reparsing the source files is quicker.
We’re still working out what to roll out in terms of geographic search features.
November 13, 2008 – 12:24 pm
As part of the Find Your MP project, a member of our team created a Generic Phusion Passenger Server VMWare virtual machine instance at Elastic Server. Don’t worry if none of that made sense.
What this means is: we use Elastic Server’s – er – service to create servers for local and cloud deployment. We use them in testing, development and production. We’re happy with what Elastic Server provides for us. Very happy. We’re even happier about the price.
We aim to release as much of our work as we legally can, which means practically all of it. This now includes parts of our infrastructure, released so others can build on it.
Why didn’t we just use Elastic Server’s default Passenger build? We tried it, but it didn’t do all we wanted exactly the way we wanted it to. We were able to tweak and build our own version and make that available publicly. Wins all round!