At the moment I try to understand how the splitter works to make the mulitpolygon code splitter compatible. I have taken the OSM dump of austria from Jan 8th and run splitter on it (with the default settings). Splitter generated 6 tiles. So far so good. To get a visual feeling I wrote a simple gpx dump at the entry point of Osm5XmlHandler.endDocument(). That makes it easy to visualize the data (e.g. with JOSM). Attached you find the files: bbox.gpx: this is the bounding box of one tile 39612875.gpx, 45656928.gpx and 456456929.gpx: three ways taken from the tile. *.png : visualization of the ways including the bbox. The splitter homepage (http://www.mkgmap.org.uk/page/tile-splitter) tells me "Tiles join exactly with no overlap or gaps." Now my questions: * Why does the tile contain points of way 45656928 and 45656929? They don't intersect the bounding box and they are contained in the adjacent tiles. For me this does not comply to the splitter homepage statement. * Why does 39612875 contain some points within the bbox and some out of the bbox? I think because the points are exported for the two other ways? * But why are the points of way 39612875 (point 474857138 and 474857142) that lie next to but in the bounding box are not contained in the tile? The described behaviour of the tile splitter makes it complicated to implement a reasonable behaviour of the mulitpolygon code on the borders of the tiles. WanMil
