State of the OpenStreetMap (horse=yes again)

Thoroughly enjoyed OpenStreetMap's conference in Manchester yesterday.

For a start, it stopped raining for about an hour in the afternoon for the first time since I moved here.

Second, as a rambler, I am in love with . Nick takes take a weekly data dump from openstreetmap, adds a contour map from Nasa, and - best of all - allows annotations on the map, such as "this path is a bit scrambly" and (I envisage) "field on the left contains farmer with shotgun".

Third, I enjoyed seeing that London motorcycle courier timeline GPS traces thing again:


.. but was particularly interested to hear Steve's comment that it wasn't actually as useful as envisaged, since the GPS traces had no accompanying metadata. Hmm, a gratifying point for metadata geeks.

Fourth, Andy Robinson talked about tagging and standardisation. Openstreetmap suffer the same problem as Quakr - synonyms and homonyms, tag soup, emerging taxonomies not emerging in the way you expected - in short, the "horse=yes" problem. Like us, they don't want to impose a standard like GML or prevent users from typing freeform, but like us, want to make some use of tagging metadata and hence could use a little standardisation. Andy had a few suggestions, including:

1. Get people to reuse terms where possible, even if they aren't a perfect fit.
2. Have a Q&A style wiki to facilitate this.
3. Use definitions from the Oxford English Dictionary if in doubt.
4. Use namespaces - Andy is suggesting STAG_BASE: for things like rivers and fields, STAG_BUILT: for buildings, canals, etc, STAG_ADMIN: for road classifications etc.

Although I like namespaces and think they can work for tag data (geo:lat is pretty popular), I'm not completely convinced they'll solve the tag soup problem. Namespaces are good for saying what vocabulary you're using, or disambiguating two keys which belong in two domains, but did we really have two kinds of horse? It's definitely a good way of categorising things though.

And all power to Andy for tackling what to my mind is the most important and interesting problem facing anyone dealing with tagging metadata - that it's a Messy Mungy Business. (See what does 'tilt' mean, anyway?" )


Finally, you just can't beat artwork like this:



Enjoy!