[Tagging] Using multipolygons to map bays in Alaska

[Daniel Koć]

W dniu 15.11.2018 o 16:01, Frederik Ramm pisze:
> Frankly, while I share your overall criticism of 'polygons is
> universally the preferred way of mapping no matter if verifiable or not'
> and 'way_area equals cartographic importance', I recently found that
> someone had added a huge natural=bay relation for the Bay of Bothnia,
> meticuously including every coastline way.


That was me.


> I deleted it here
>
> https://www.openstreetmap.org/changeset/63455985
>
> but not without attracting questions as you can see, and I fear that
> with osm-carto as it is, this object will rise from the dead sooner or
> later.


I wanted to know what was your arguments, because I found it
questionable action - especially without even contacting me. Thanks for
writing them. I wanted to answer, but had no time lately. Fortunately
David started this thread and you have bring the topic, because
discussion list is a better place for such a generic subject.


> I still think that it is ridiculous to slap a ton of multipolygons onto
> the coastlines (since, as I write in that discussion, many bits of water
> are actually part of various bays of different scales), and a coarse
> "labelling polygon" like the one discussed here would seem to me to do
> *less* harm.


Words like "ridiculous" do not help to solve any issue, so let's not use
it in discussion, please.

For me rough polygon - however it's something new for me - is indeed
doing less harm than claiming that a node is a valid approximation of a
complex area. Especially such a big one, where even from the globe scale
(z0) is still visible as an area. It's like claiming that we can
estimate country as a node (many of them have similar size).


>> Long story short:  My suggestion is and has always been to map bays with 
>> nodes in those cases where this - together with the coastline - 
>> perfectly documents the verifiable information available on the 
>> geometry of the bay.


But this is oversimplification.

When you put a node, you claim that it's inside the bay area, isn't it?
But since you claim that you don't know where the border is, you also
can not verify it for sure. It's not any better when it comes to proving:

https://www.openstreetmap.org/node/5978404937#map=5/61.669/26.609

The coastline also does not help, because you still don't know which
part of the coastline is also a part of a bay (that could be a whole
continent for example, which makes the bay to be so called "global
ocean" - definitely not helpful).

Also the node location is random. Why not 500 km on the north-east or
south? With area at least location is sure (you won't move it anywhere
else), at the same time the shape and size are much closer to reality.

So with a node you claim that you don't know anything about the borders
but you act like you know them quite well.

I agree that this is intuitive idea that "node is always better", but
such intuition has serious limitations. It's valid only when you don't
care for details for now and plan to make a sketch to be improved with a
real geometry later on.


-- 
"Excuse me, I have some growing up to do" [P. Gabriel]