Yeah I know of that github project and I'm thinking of that an approach of having small fuzzy areas in the main database, and huge ones in a separate might be the way to go.

One reason to have big names separate and not the small ones could be that the big names will be "completely" mapped almost right away and don't require crowd sourcing. However the small ones, as an example we have say 5-10 of these names per 10x10km square in Sweden, do require crowd sourcing from regular mappers.

But then I ask myself, if we can have the small ones, why not have all, also the big ones, in OSM? We have already big scale information in the database. The clutter thing I think is just a tool issue which is already solved in JOSM (filters) and can be easily solved in iD.

Or we could go the opposite way and move all fuzzy areas to an external database, also the small local ones. It's sort of a conceptual way to create it as a separate layer. I'm not against that from a technical perspective, but I'm worried if this data is not included in the main OSM database it's a big risk that it won't be available and won't be used by OSM-Carto and then mappers won't be motivated to contribute so we won't get the necessary crowd sourcing going.

(I've heard that some think fuzzy areas is "mapping for the renderer" and that's the reason we don't really move forward on this issue. Unfortunately I don't remember the exact reasoning behind why it would be mapping for the renderer so I can't recreate that argument here, but I guess someone can fill that in. From my perspective I think fuzzy areas is the exacty opposite to mapping for the renderer, as the mapper just give information of name and rough size of an actual fuzzy area, and makes no decision on label placement or size. Sure one can misuse it and say make a fuzzy area much larger than it should be to make the renderer draw a large text just for the sake of it, but that's just regular misuse and something we need to relate to for all OSM tags and features.)

/Anders

On 2020-12-21 11:03, Janko Mihelić wrote:

The fifth alternative is move the big areas to an outside repository:
https://github.com/dieterdreist/OpenGeographyRegions

This might be a great alternative until we find a better solution. And there might not be a better solution.

Janko

pon, 21. pro 2020. u 10:22 Anders Torger <[email protected]> napisao je:

Next question.

In the mountains we have an number of named plateaus. There is a tag
proposal for natural=plateau, but just like with natural=peninsula and
similar tags there is an underlying question that we really need an
answer to first: should we have fuzzy areas or should we not?

Plateau, peninsula etc are naturally mapped like an additional low
detail fuzzy area polygon on top of other land covers. My opinion has
been made clear in other threads: I think fuzzy areas on top is an
elegant solution for naming nature and something we should have. I think
the cluttering issue can be solved with filters, but as these will be
used in low numbers to start with I think cluttering will not be an
issue for some time to come so it's something we could look into later.
In any case that's a tool issue, not a database issue.

If we don't want fuzzy areas, an other alternative is to have these as
named points, (previously often made as "place=locality"). I think that
is okay too, but then we need size classification on them like we have
on residental isolated dwelling/hamlet/village etc so the renderer have enough information to know how large to make texts and which zoom level
to show them. Having the same level for all names doesn't work.

Fuzzy areas has the advantage of solving the text size automatically
(not a mapper decision), and gives freedom to the renderer to place (and even shape) the text. Fuzzy areas also scale well up to huge sizes (like the Sahara desert) if we want that as well, which point text doesn't in the same way. We could decide to have fuzzy areas over a certain size in
an external database too. I'm not super-stoked over the external
database method though, as I think then it risks becoming like
elevation/contours is today, ie not generally available and with varied
quality.

A disadvantage is that fuzzy areas have limits in verifiability and it
arguably requires more knowledge/judgment from the mapper than roughly
placing a point. On the other hand, optimal point placement also have
cartography and verifiability issues. The underlaying issue here is of
course that these type of names have never have defined borders and
never will, but I think we cannot continue to pretend that they aren't
relevant for a database mainly used to generate maps. We need to
represent them in some way.

A third alternative is not having names of this type at all. While I
just said that it's not the way to go, if someone still has that as a
clear opinion please make that clear rather than just point at
disadvantages of every suggested solution without coming up with an
alternative. We know there are disadvantages and no solution is 100%
perfect, but sometimes there's a higher goal to fulfill.

The fourth and current alternative is leaving the question undecided,
with some fuzzy areas active (bays and straits), some not rendered
(peninsula), and passively see how it plays out in the coming years (or
decades!). It's the simplest alternative, but as a mapper and OSM end
user I hope we can make some real progress now.

Worth mentioning is also the alternative to make a fuzzy cutout of the
dominant landcover and name that. I've done quite some forest naming
that way. However it's quite complicated and time-consuming to make
these cutouts (complex multipolygon editing), and it only works well
when the name is actually tied to the landcover as such, eg the name is on the forest, not a forest-covered peninsula or plateau. While I think it's okay to mix this cutout naming method when it works, and use fuzzy
areas on top when that is required, I also think a viable option would
be to name forests with fuzzy areas on top as well, but then we need a
specific tag (or tag combination) so the renderer knows that it
shouldn't make landcover rendering for that.

I'd like to at least know where we are headed. I could use a tag which
is not yet rendered, but it would be nice to know if the information
will potentially ever be used, or if I'm maybe just wasting my time...

/Anders

_______________________________________________
Tagging mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/tagging

_______________________________________________
Tagging mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/tagging
_______________________________________________
Tagging mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/tagging

Reply via email to