I repeat, again…
It’s wrong !
I’ll even give you an example and I’ll stop there…
This island is inside a natural=bay multipolygon without being a member of it and… it is visible on the renderer.
If this is the case for natural=gulf also, then you are right, I have no more worries!
I propose an honest deal: I take out of the Gulf of Ajaccio relationship all the islets. If the natural=coastline still appear then I apologize and I recognize my (big) mistake. Otherwise, we replace the Gulf of Ajaccio by a node.
Do you agree ?
In all renderings?
In the rendering of the islets area marked as natural=coastline.
Because of a reason I don’t understand, Olyon included all these islets in the natural=bay relationship of the Gulf of Ajaccio. I thought it was necessary for them to appear. Why did he do this? Already without it this relation is a monster.
The official data of the coastline in France (an Corsica) is evolving and I would have liked to be able to update it without unnecessary difficulties.
Moreover, the coastline toponyms require for their understanding the representation of these islets. For example i setti navi, the seven boats, are seven islets, not two as today on OSM!
These tangible data seem to me more important than an artifice of presentation, but I don’t want to offend anyone.
From the point of view of someone rendering OSM data, as far as I’m concerned the most important thing is that any natural=bay polygons only duplicate coastline-enclosed areas.
We have had the situation in the past where an inland bay (not coastline-enclosed) and sea bays (coastline-enclosed) have both been mapped as natural=bay. This makes it really hard for renderers to know what to do. Not everything chooses the same rendering order as openstreetmap-carto.
2 Likes
I feel it’s very interesting but I’m not sure I understand it well because my English is poor. Can I rephrase you like this ?
The most important thing is that the natural=bay polygons consist only of natural=coastline areas.
Does this mean that depending on the rendering order chosen by the programmer the islets that are natural=coastline but not part of the polygon are hidden or not?
If so, then there is a real problem of complexity because there are too many islets to include them into the polygon fluently. In practice this forbids any evolution under the risk of losing data.
Put simply, someone developing a stylesheet needs to consider “do I need to fill a natural=bay polygon in blue?”.
If all natural=bay polygons just duplicate the coastline, then the answer is no, because the coastline is already filled. 
If some natural=bay polygons cover areas not covered by the coastline, then the answer is yes, because otherwise there’ll be missing water areas.
But if this is the case, and islets aren’t included in the natural=bay polygon, then renderers are at risk of colouring islets in blue. (And yes, I’ve seen this happen.) 
Conclusion: make sure your polygons only duplicate the coastline. (Or just use nodes.)
3 Likes