[Tagging] building and man_made -- was: Re: Feature Proposal - RFC - (artwork_type=maypole)

[Georg]

Dear all,

on 2021-10-21 13:40, Mateusz Konieczny wrote:
> man_made is not for "structural structure like a building"
>
> we use building=* for that

When reading this bold claim I thought it's on very, very thin ice in
the sense our definitions for man_made and building are IMHO somewhat
unsuited to *clearly tell apart* between the two.

So I looked a little into that topic; some details (you may skip this
section): https://wiki.openstreetmap.org/wiki/Key:building tells "a
building is a man-made structure with a roof, standing more or less
permanently in one place". Okay.
https://wiki.openstreetmap.org/wiki/Key:man_made lists e.g.
* lighthouse. It even contains the term "house" in the name and people
use(d) to live in many of them! It has a roof and all characteristics of
a building. Still, we don't have a building value for them.
* dovecote. Even the picture in the wiki shows an object having a roof
and looking very much like a building and matching the definition of a
building. So why is it in man_made? Moreover, it's for an extremely
similar purpose like a cowshed, stable or sty - but still, it is
classified as a completely type of entity...hard to see this as consistent.
* bridge. For bridges with a roof - excepted for skywalks - we don't
have a building value despite they are meeting the definition, like e.g.
https://vietnam.de/wp-content/uploads/sites/4/2016/04/VIE_2011_1JEH_Bruecke_2618_5d2-2.jpg
  Same story with goods_conveyor, pier, etc.
* crane. https://de.wikipedia.org/wiki/Oestricher_Kran has a roof, looks
like a building and fulfills definition of building. Again, we don't
have a building value for them. Same for communications_tower. Even the
picture in the wiki shows a tower having a roofed & enclosed story and
thus meeting building definition, many other communication towers like
https://de.wikipedia.org/wiki/Berliner_Fernsehturm fulfill the
definition of a building even more clearly, but we don't have a building
value for them. Similar for  gasometer, monitoring_station, observatory,
offshore_platform, reservoir_covered, silo, storage_tank,
street_cabinet, stupa, watermill, etc.

Conclusion:
For all of these man_made values, we do not use building=* despite
building definition is fulfilled. It's such an amount of examples that
IMHO, we can't just ignore them as exceptions of the rule, but we have
to acknowledge it's a pattern questioning the whole rule (so not only
your bold claim which triggered my reflection). To me, it seems we did
really mess up these 2 keys. IMHO changing the definitions so they do
exactly match the current values _and_ be easy to understand will be
impossible. Much more feasible would be to move values that always
fulfill building definition (like gasometer and reservoir_covered) from
man_made to building, and copying values which may or may not fulfill
building definition (like crane). That would also result in more useful,
semantically more rich data, because a man_made=pier will obviously be a
simpler example than a building=pier (e.g. https://bit.ly/3wqHgP8 versus
https://bit.ly/3kfPhla)
=> Shall we dare to clean up? If yes, by moving/copying values or different?

Greetings, Georg