Computing, GIS and Archaeology in the UK

The geospatial world rightly got quite excited over the last few weeks with the discovery that Geofabrik are offering openstreetmap data in shape file format. You have been able to export openstreetmap data to postgresql (and from there to shape or whatever takes your fancy) for some time, but this makes the process of really using the data far easier.

However, I got to wondering exactly how useful it would be as a real-life background dataset (in the absence of anything affordable from the Ordnance Survey or similar), when you would need to repeat the process ad infinitum to keep the dataset up to date. I remembered that once upon a time there was an openstreetmap wms server, but that it was taken down several years ago.

A quick google search, however, shows that the Geofabrik crew are actually in the process of putting up their own wms server and it’s in testing at the moment. This is fantastic news, and I for one look forward to more information about their progress!

Exposing our data using WFS has certainly prompted some debate, which is good! Particular thanks go out to Andrew Larcombe, who put together a great OpenLayers page, with some nice ajax touches, and a clever mechanism that clusters nearby sites together and reduces the load on the server.

We had a few requests to release this data by wms as well, so here is the request for that. The process of doing this has thrown up some challenges, which I think are mostly down to my ignorance than anything else!

Firstly, as I understand it, if I want to expose the data via WMS using mapserver, the projection of the map has to match the projection of the layers, unlike with WFS where I can re-project the data from British National Grid to WGS84. That means that I need to convert the underlying shape file rather than relying on mapserver, and as I update the data it will need to be redone each time. Once we go over to Postgresql this will be less of an issue, but the fact remains that we will always enter the data using British National Grid coordinates.

Secondly, rendering the large number of points (thousands) via wfs has proved slow. No surprise there. How to get around this though, and still maintain the flexibility that wfs engenders? Andrew’s clustering mechanism might be the way forward, if I ask him nicely…

Thirdly, I have begun to put together a mapping interface for this myself. It’s not pretty at the moment, and subject to change, but at least it’s a step forward. It also uses Yahoo and Microsoft Virtual Earth data for the background mapping, which goes against our plans for a totally opensource approach. However, I started off with NASA Worldwind, and found it very slow. I’ll admit that I have to experiment with tuning all the different components of the map, so I’ll revisit this later.