[Tagging] Feature Proposal – RFC – natural=peninsula (Was: Feature Proposal – RFC – place=peninsula)

Martin Koppenhoefer dieterdreist at gmail.com
Thu Jan 10 08:50:20 UTC 2019



sent from a phone

> On 10. Jan 2019, at 08:54, Tomas Straupis <tomasstraupis at gmail.com> wrote:
> 
> Sorry, I did not get it. How saving only vertexes is better than
> having a polygon (made out of those vertexes)?
> 
>  Full geometry is required to be able to calculate label positions on
> all scales.


The full geometry would not have to be stored in the db as a polygon , it is sufficient to have the information to be able to create it. This would make evaluating these things more expensive therefore could be done not by default only by request (eg in osm2pgsql), for fuzzy border objects for example it doesn’t make sense to draw a precise border anyway. For a peninsula for example (not a fuzzy border object, at least not completely) it would be clear that the sea is not part of it, so in theory with just 2 coastline nodes and a definition how to order them, the iberian peninsula would be sufficiently defined (for the land border we could look at the watershed? Ridges etc.?)
A problem with this approach could be to find the relations if you wanted to edit them, but this could be solved by a search index (nominatim etc), and it would make it more complicated to understand what happens why (understand the data structures).

Geographic regions on the land are more difficult because they often don’t have a precisely defined border. This also means you do not need to know it, it would be sufficient to have a coarse idea about the extension/size and shape, so you could decide when and where to draw labels or show the region as a search result. A new datatype would be needed for this, e.g. by telling a few things that are still clearly inside and clearly outside the region it would be somehow possible to “guess” the rough geometry in these cases.


Cheers, Martin 




More information about the Tagging mailing list