On Mon, Jan 11, 2010 at 11:01 PM, Tobias Knerr <o...@tobias-knerr.de> wrote:
> > I disagree. The most important thing is that different people using the > same tag use it for the same kind of thing. > > Its a matter of seconds to replace all amenity=vet with > amenity=veterinary once you have noticed that there are two tags with > the same meaning. It would even be trivial to simply support both in a > renderer if cleaning up the database isn't an option for some reason. > > However, if you find out that there are 627 contradicting > interpretations of "cycleway" and you cannot find out which use of the > tag follows which interpretation, fixing that mess is much harder. > Yes, I think you're right. Although I would slightly narrow your definition to be something like "the most important is that different, indistinguishable uses of the same tag refer to the same kind of thing". That is, if all uses of highway=cycleway in Germany mean one thing, and all uses of highway=Cycleway in Russia mean another, then that's still not too bad. > > That's an easy case, because there isn't really a "better" solution > here. Many tagging scheme alternatives are a bit more complicated, > though, such as addr:street tag vs. associatedStreet relation. Yeah, mapping amenity=vet onto amenity=veterinarian is obviously a simple example. When clusters of features are represented in single tags, it can get much complex. IMHO, this is an argument for small communities defining their tags in ways that make sense to them, using them in that sense, then attempting to align the tags with other communities later. But not everyone agrees with this kind of approach. > In that > kind of situation, early normalisation might easily lead to mandating > the solution that would, after some experimentation, have turned out to > be the inferior choice. > > Yes, early normalisation can cause big problems. In many parts of OSM though, I think we're well past that stage, and normalisation would be well and truly "late" now. Steve
_______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk