On Wed, Mar 06, 2013 at 01:48:58AM +0100, Martin Koppenhoefer wrote: > Am 05/mar/2013 um 15:41 schrieb Jochen Topf <[email protected]>: > > > We are also proposing to remove the huge natural=glacier multipolygon that > > covers large parts of Antarctica. It is quite unusual to tag the "default" > > land > > cover this way (there is no "desert" multipolygon covering half of North > > Africa > > > probably it would be there if there was more activity in Africa in OSM. > Usually we don't accept deletions because a feature is not yet mapped > elsewhere. I am opposing the idea to remove the "default" landcover. > > > > or "forest" multipolygon covering Siberia). The problem is that every time > > somebody adds some land cover information (for instance natural=rock for > > some > > places where there is no ice cap) it has to be added as inner ring to this > > huge > > multipolygon. Sooner or later this multipolygon will be too big to handle > > properly. > > > In this case we could then split the relation into smaller parts like we also > do with forests or other huge mp. > > I generally like the idea to import data for the Antarctic and am not against > deleting very rough coastlines or other data there (with the intent to > replace it with more accurate data by import and with the old data coming > from previous imports), but you shouldn't delete "correct" mapping and you > should pay attention to the (maybe existing) manual refinements and mapping > in the area (I.e. keep it). It is a really huge region and while there might > be a lot of fun and none sense objects there might also be some treasures.
We will certainly keep manual refinements. There are very few places where existing data is good and we will keep that of course. But the situation with the ice shelf multipolygon is different. (There are actually two multipolygons here covering the parts of Antarctica east and west of the Ross Ice Shelf.) There is not really any data lost if we remove them. Basically all those multipolygons do, is tie all coastlines together into these two relations. Except in those places where somebody improved the coastline and the multipolygon still references the old, bad coastline, so it is definitely wrong there. Oh, and there is one inner ring with natural=rock. The Multipolygons only go to the 85°S line that is the border of what can be shown on Mercator maps. So the data is definitely wrong and it is not easy to make it work in all maps. You can see the kind of rendering problems we get here: http://polar.openstreetmap.de/antarctica/#zoom=5&lat=-87.70367&lon=-137.08148&layers=BT I think these large multipolygons make more problems than they are worth. With or without them, you can't render Antarctica properly without specialized software and major data fiddling, because of the problems at the -85° line, South Pole and -180/180° line. Jochen -- Jochen Topf [email protected] http://www.remote.org/jochen/ +49-721-388298 _______________________________________________ Imports mailing list [email protected] http://lists.openstreetmap.org/listinfo/imports
