On 18/11/2010 20:18, Dave F. wrote: > If key tags don't mean anything, why have them? >
It's a quirk of the way tagging works more than anything. We need key-value tags for properties like name=* and oneway=* and there's no point in having two separate tagging systems for "class" type tags and property type tags. I believe, once upon a time, shortly before I got involved in OSM, there was a class=* tag, but since you always had to have a highway=* with each class=highway, and you couldn't have more than one class=* value, it was decided it wasn't very useful after all, and it stopped. In an ideal world, you could have more than one value per key, and we'd have (for example) class=secondary_highway or class=theatre, or some shit like that. This makes the whole "which key?" problem go away. The few occasions where you need to assign two separate classes to the same feature make that problematic. There's a certain amount of utility in being able to grab all highway-related features in something like XAPI or Osmosis using highway=*, but for less clear cut or overlapping keys, like tourism=* or historic=* it just doesn't work as well. You could use duplicate keys (tourism=theatre, culture=theatre), but you could just as easily use TagTransform to solve the same problem. So we have keys for class tags mostly because we *have* to, and because for certain types of features it helps with identifying a general grouping for the tag, but not all. Most of all, changing foo=bar for baz=bar achieves absolutely *nothing*. Hope this helps -- if I've missed a point let me know, because this is turning into a good draft of a post for a wider audience. Jonathan -- Jonathan (Jonobennett) _______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb