Users have started complaining that Corine landuse imports are overlapping water areas in Finland. If the Garmin first draws the water and then the other landuse (forest, farm, etc.), the waters will be hidden and the map will be badly useable for navigation. What if we separated most polygons to a separate map layer? For non-terrain uses, I cannot see any use of landuse polygons, possibly apart from building polygons. One reason why polygons are currently needed is the --add-pois-to-areas option. That problem could be solved by implementing a polygon style file action 'add_poi'. If we start splitting split stuff to separate map layers, another good candidate would be power lines, pipe lines and such. They are also mostly useful for navigating in terrain, I would guess. There could also be a separate layer for boundary=*. But, if we start adding layers, it would be useful to generate multiple map layers in a single parsing pass. Otherwise the parsing time is going to dominate. Some time ago, a style that would generate no output would require 2.5 minutes or 3 minutes for processing the Geofabrik extract of Finland (about 90 megabytes in osm.pbf format) on my dual-core system. Also each routes-* style needs about this much processing time, because they do not generate much output. About half a minute per layer could be saved by disabling the processing of multipolygon relations, but I guess that we could reduce the processing time to 2 minutes for all routes-* layers if we could generate multiple *.img files from the same input tile using some special style action syntax. Best regards, Marko