[Tilesathome] close-areas producing very long ways -> inkscape crashing

[Matthias Julius]

Frederik Ramm <frederik at remote.org> writes:

> If we want to use this method for all other types of areas - lakes, 
> woods, landuse areas, natural reserves - then we would have to find a 
> working equivalent to the "tile index" we now have. Let's take the 
> example of natural reserves as they are often quite big. Whenever you 
> render a tile that does not contain a natural reserve boundary, you 
> would have to somehow find out if the whole tile is inside a natural 
> reserve. For this, is might not be so easy to create an index - and much 
> less so for other areas.

This problem is with the current method just the same. If the bounding
box is completely within an area the renderer won't know about it.

There was talk about a server based oceantiles system. If that is
implemented it could support more than three values. The data could be
collected automatically. A script could request all areas with
natural=water from the XAPI and figure out if any of them completely
encloses a z12 tile. The same of course for other potential areas.

The problem of huge ways for areas could also be solved by using
relations. One could break up that way and make all the fragments
members of a relation with natural=water for example. That just needs
to be supported by the renderers.

Matthias