The last time I was at an OSM London meet-up was four weeks ago. Feels like a long time. We were in the Mulberry Bush, and I seem to remember it was a good gathering.

Yep. Lots of people around the table. I have some notes. Some interesting conversations as ever, so, even if this is a bit out of date, here's what we talked about:

This meet-up took place the day after the Christchurch earthquake. TomH was, at the time, on holiday in New Zealand, and for few hours there had been concern that our chief sysadmin and rails developer might have been earthquaked upon. But no. Firefishy saw activity from him on the server a few hours later :-) Since then there was some crisis response mapping efforts coordinated on the Christchurch earthquake wiki page (And of course various other disasters have occurred since then. See my HOT blog post)

Also just as we were meeting last time, Enfield had reached 100% on the OSM and OSL differences analysis. The only London borough to have achieved this so far. Congratulations to Enfield. Do we have Steve8 to thank for that?

Bing imagery is better than the old Yahoo... or is it? I had assumed that throughout the globe, at any given point, bing had equal or superior resolution imagery to the Yahoo stuff we used to use. Well we still have access to Yahoo!, and it's probably a good thing because it turns out there are patches of high-res coverage which bing does not have. For example the handful of cities covered in by Yahoo! imagery in Pakistan... Bing has nothing there. That's bad news for editor developers. Because of the way Yahoo requires access to imagery via their javascript library, the code for accessing within editors is fudgetastic kludgefest in both Potlatch and JOSM. It's also the reason Yahoo support was always a big pain to install in JOSM. Bing imagery is more straightforward, but given the situation in places like Pakistan, we probably shouldn't drop support for Yahoo! just yet.

Speaking of imagery coverage, should the boundaries of areas of coverage be sketched into the map data? It's clearly a violation of the on-the-ground principle, but it has some very practical uses for mappers, especially for wide-ranging imagery sketching (which can be controversial in itself). Imagery boundaries were used to good effect in the early days of Haiti response, to help keep track of new sources. However another disadvantage, is that our natural instinct to share URLs to these boundary elements, can cause a problem for the server. This is something I noticed, purely in terms of the length of time it took for the server to respond, when looking at the Pakistan imagery a few months back, and so when I made my OpenLayers display of coverage I implemented a caching mechanism. PHP keeps a copy of the relation XML, and gives that to OpenLayers instead of hitting the OSM server. This was partly as a way of just trying to make the display work quicker. I'm not really sure to what extent the server actually struggles to cope with these things, but it came back into discussion recently (here's how the recommendations stand at the moment) and in the pub Wyndale was suggesting that a MediaWiki Extension could perform caching for relation links, similar to the php code I've already written there. Would be easy enough to do I think. Conceptually it's a little odd though. Wiki mechanisms to cache content from API? Good idea? I'm not sure.

We had along Ritchie, who is a developer for SABRE the Society for All British and Irish Road Enthusiasts. He also seems to be an out-of-copyright Ordnance Survey enthusiast. OpenStreetMappers tend to be quite enthusiastic about those kinds of things, so he fitted right in.

Dan Karran was telling us how the Isle of Man mapping has taken steps into the murky maroon coloured world of building mapping. Probably a more attainable target than London.

We talked about matters of search and routing. Some new hardware had been specced out for running this kind of thing as an OpenStreetMap service, but currently (or currently back then) the shiny new machine was just forwarding requests to a MapQuest nominatim deployment. Having an OpenStreetMap overseen service might be better for the project because MapQuest have funded mostly U.S. oriented development (search test cases which work well with TIGER data) There was talk of routing using a "Contracting hierarchies" algorithm, which would need some hardware tuning for this RAM intensive processing. As always there was the everlasting confusingly circular discussion on whether rendering and searching systems should be geared towards users or mappers.

And speaking of old discussions, we chatting about temporary mapping (temporal, 4th dimensional, historical, life cycle mapping... whatever you want to call it) There's an old wiki page which summarises various tagging ideas, including the stupid ideas such as disused=yes. If you think about users of the data, this is obviously a fairly disastrous breach of the "principle of least surprise".

We chatting about XAPI, which is an ongoing and much requested development priority. Firefishy speculated that some XAPI users may not realise that osmosis offers very similar querying power, which you can run locally. Shifting processing onto a local machine may be advantageous or disadvantageous, but certainly you can do all the same type of things: extracting data by bounding box and by different tags. So an idea to help with this might be a little tool which translates any given XAPI query URL to an osmosis command.

Andy and Jon Burgess were sharing Mapnik know-how with me. This time we were talking about osm2pgql and how it usually outputs to spherical mercator projection, but Mapnik can convert (and other details which I didn't write down)

I wrote down "contributions visualisation". Can't remember what this was about. Maybe Matt's OWL (OpenStreetMap Watchlist). Maybe Pascal Neis' "how did you contribute to OpenStreetMap", or other ideas.

I also wrote "Keeping up with translations". Can't remember which particular aspect of translation that conversation was about.

Finally I chatted with Andy about the upcoming Hack weekend (Saturday 2nd and Sunday 3rd of April), and about ideas to try a "workshop" format. There will be the usual hacking going on throughout, but this is an idea to present a more welcoming face to any technical folk who want to try and get more involved in OpenStreetMap development. Ideas for topics include Potlatch 2, The Rails Port, and git. To me this sounds like a great idea, mainly because I can learn a great deal myself on all those topics. The hack weekend is already looking pretty popular with familiar names. It's going to be big one! But we should try to think of ways of attracting some new hacking talent too. Know anyone who may be interested in the workshops? Ideas welcome. All the details and sign up: 'London Hack weekend April 2011' on the wiki.

That's all pretty soon, but not as soon as the Last winter pub meet-up this Thursday at the Monkey Puzzle!. That's the pub in Paddington/Edgeware Road area. We started the winter season with a trip here. It seems fitting to end it that way. Don't miss it! It's the last time I'm ever going to ask you to sign up on the 'London/Winter 2010-2011 Pub Meetup' wiki page!