[Tagging] Using multipolygons to map bays in Alaska

[Christoph Hormann]

On Sunday 18 November 2018, Eugene Alvin Villar wrote:
> [...]
>
> As a sort of compromise at least for bays (gulfs, inlets, fjords,
> coves), how about we just map them as a single way across the mouth
> of the bay and not as a way-polygon nor type=multipolygon relation?
> And then we set the direction of the way such that the right-hand
> side of the way points to the bay-side (just like the right-hand side
> of natural=coastline ways point to the seaward side).

While this obviously has the same verifiability issues as the polygon 
drawing in general (you already say that) this is actually a great 
demonstration for the core of the problem.

It can be explained with the following formula:


polygon mapping of bays/straits using the coastline as components

*equals*

coastline data already mapped anyway

*plus*

completely subjective data about a non-verifiable limit of the 
bay/strait


Mapping only the last part should satisfy all those who disagree that 
there is no additional verifiable information in the polygon mapping of 
bays (because whatever is in there would be contained in this form of 
mapping - see the above formula).

It will however likely not satisfy most because:

* The "map designers who want to outsource label drawing to the mappers" 
and "mappers who want to draw labels" will have difficulties acutally 
performing above addition (because as mentioned: coastline data is not 
in the database and the operation to select and assemble the data as 
needed is not cheap).
* The "everything is to be mapped with a polygon" proponents will not 
like this because it is not a readily assembled polygon, just a 
component of it, therefore insufficient for a purist.
* The verifiability proponents (like me) will dislike adding 
non-verifiable data to the database but this is much less harmful than 
the polygon drawing so i clearly see it as the lesser of two evils.  It 
nicely separates the verifiable data from the non-verifiable data so it 
definitely is the most acceptable form of non-verifiable mapping in OSM 
(if there is such a thing).

As a data user - while this is more difficult to process than a node 
based mapping it would be manageable.

Long story short:  What i like about this proposal is that its rejection 
will bring clarification on the motives for the different opinions 
pursued here.  What arguments you have against this suggestion will 
decide which of the above groups you belong to. ;-)

-- 
Christoph Hormann
http://www.imagico.de/