We rounded off the London winter mapping meet-ups with a trip back to the Monkey Puzzle in Paddington. I forgot to take any photos! Too many interesting conversations going on:

Simon Nutall, one of the CycleStreets developers, was visiting from Cambridge. Naturally we got talking about routing. With it's triple route response (quiet, fast and balanced route), cyclestreets is a sophisticated OpenStreetMap based routing system, but I was surprised to learn that it doesn't heed turn restriction relations. These don't actually occur all that often, and it always seemed doubtful to me that they are desperately important to capture for routing. This confirms it. Note that most instances of what you might call a "turn restriction" are mapped more simply, as one-way streets, without the use of relations. It's good that we have a mechanism for mapping the more complex restrictions, but it's not an essential part of mapping.

Having said that, Simon told a rather alarming story of how he once received a fine from the council for performing an illegal turn on a bike! They posted him evidence captured on camera. todo: link to his blog post about it?

We also had livingwithdragonsMarlerTheGregmeisterGregory in attendance. Down from Durham on business. He has a geojob! Congratulations to him on that.

Richie from Sabre was there. We chatted about amusing sign post screw ups. This one in Wales is my personal favourite.

JennyH was there. She's just left her job at nokia. A lot of people are quitting nokia's burning platform these days it would seem. So she'll be on the look-out for new opportunities I guess. She said she's joined the London hackspace, so we can watch out for her on the webcams.

Derick is moving house. This is good news. He's created his own building blob in North West London. If all of our pro mappers would just move house more often, we'd have the world mapped in no time!

We talked about "geovation". This is Ordnance Survey's scheme to boost geo... innovation. See what they've done there? There's prize money to be won. Not really massive get rich quick amounts of money. The prize is mainly whatever credibility and kudos ordnance survey have to offer.

I told people about a nifty thing Twain has come up with. OpenStreetMap rendering looking like Ordnance Survey. Isn't that weird? ...and awesome. It's available worldwide too (at some zoom levels) If the Ordnance Survey had ever made a map of Paris it might've looked a bit like this:

This idea of style matching seems like a powerful way to demonstrate the flexibility of OpenStreetMap rendering software. Or in some ways it nicely de-emphasises the flower-arranging aspects of it. Ordnance Survey fans always harp on about the map styles. This is a great way to make people think about the deeper data issues.

I've often thought that an interesting stylesheet for mapnik would be colour choices and line widths matching those of osmarender's defaults, or vice versa. Making the broad style choices similar would be a good way to reveal the real pros and cons of the rendering technology. I notice this trick has not gone unnoticed by mr maperative who has a very Mapnik-alike style available:

QualityStreetMap version 2 is another neat thing. Or at least I think it has some potential to be a neat thing. I was trying to describe it in the pub. It's a mapping coordination tool relying on people assessing quality of map tiles. The neat thing is essentially the nice javascript interface for the tile selecting (including ctrl+click). I think the idea is to add tags to tiles to represent your assessment of a tile being 'complete' on that tag. I'm not entirely sure. Clearly it needs to be made a lot more obvious, and isn't anywhere near gaining the critical mass it needs to be useful, but I can imagine this tool consuming the output from Andy's rapid ass to get more assessments done quickly. This then becomes the results view and tool for pro-mappers to tweak the assessment results. (e.g. mark tiles where they've made improvements).

We talked about the database. The DB server (smaug) is feeling the strain again these days. The proportion of the time spent maxing at 100% is on the increase. An SSD cache was fitted, but disappointingly it made no difference. So other scaling solutions were discussed, including partitioning the database so that GPS points are separate from the rest of the rails app data.

The number of nodes in the database recently passed the 1,000,000,000 mark. That's one billion nodes! So that's also around about where we're at with the maximum node id. 32 bit systems have a maximum signed integer range of -2,147,483,648 to 2,147,483,647 so these node ids shouldn't be too difficult to store just yet. Unsigned gives us more space. but it's fairly common practice to use the negative numbers to denote something else (nodes which didn't have server side id's assigned yet) so when we reach >2 billions, some systems/software will be hitting problems. If any developers used the number range >1 billion to denote something else... well that code just broke :-)

If that's not in depth and techy enough for you... well you'll have to ask the people who know more about this stuff than me. A good chance to do that would be the Hack Weekend this weekend. This is really shaping up to be a big gathering of chief OSM hackers, but don't forget we also have a plan to do workshops for anyone who'd like a gentle introduction to it all. Sign up on the page there, even if you're a "maybe", or you wont be able to get into the building