As a friend and I were adding the 57th tag to a climbing gym in Belgium, I wondered what the element with the highest tag count is. I couldn’t find such stats (which may be a good thing), so I downloaded belgium-latest.osm.pbf from Geofabrik and wrote a simple Python script that uses Pyosmium to do some counting for me.

Without further ado: the Belgian record for highest tag count is… the relation for Belgium itself actually, which currently has 491 tags. The non-relation with the highest tag count is, boringly, the node for Belgium, with 288 tags. Next up are Brussels (156 tags), the Council of the European Union (79 tags), one particular section of the River Meuse that somehow got its name mapped in 57 different languages (65 tags) and the Irish embassy (also 65 tags). Next up is the first element that has a lot of tags not because it’s just flooded with languages! This maritime beacon north of Antwerp in the River Scheldt has a respectable 63 tags to describe all its lights. Our climbing gym is not far off from this one, and has a lot more diverse information in its tags I’d say.

I noticed that the relation for Belgium also has a high version number, it’s at its 1043th revision. That prompted me to take a look at version numbers too. But 1043 isn’t even close to our record, which goes to the superroute relation for the E40 (version 3141). Granted, that’s international. The version record for a purely Belgian object is the hiking route GR 126 (version 1103) from Brussels to Membre-sur-Semois. Just like with tag counts, I find it more interesting to look at non-relations here, though. There the honour for highest version goes to one of the outer rings of a farmland multipolygon south of Mons, which is at version 277. Funnily enough, in contrast with its senior version number, it almost has no tags to speak of, only a source!

And should I even say… If you’re now inclined to go add tags or bump versions just for the sake of it, don’t. If I ever do a follow up on these stats and see that you’ve done this anyway, I will flat out ignore your object in future reports (yes!) and generally be very cross with you.