I can't comment all details, some general ideas. Maybe someone else will step in?
> Jaak Laineste wrote: >> > I have now done so for admin_level=8 in Hiiu, Saare and Pärnu maakond. >> > [I'll revert that if agreed here] >> How does it look like on Mapnik, can you give permalink where you have >> defined it so? > > How about (some special examples): > http://www.openstreetmap.org/?lat=59.027&lon=22.7512&zoom=13&layers=B000FTF > http://www.openstreetmap.org/?lat=58.202&lon=22.3945&zoom=12&layers=B000FTF > http://www.openstreetmap.org/?lat=58.35&lon=24.5058&zoom=12&layers=B000FTF > > And some counter example (only having coastline+relation membership with > admin_level=8 -- not modified by me): > http://www.openstreetmap.org/?lat=58.9487&lon=23.5454&zoom=14&layers=B000FTF > (Some islets having problems at the moment ... will fix that) In visual rendering the arbitrary border in sea looks better option. But technically (geoinformatically and legally) this is just cartographic hack. Paper map makers have freedom to use a lot of visual tricks, like drawing fake borders, they also move POIs (and also roads, rivers etc - depending what is more important) to fake locations to make them visible better. In digital databases like OSM you cannot do it. Or if we really need to do it, then it must be correspondingly tagged (fake border, fake location etc). IMO the proper way to solve it would be to have Mapnik rendering fixed: if line is both city border and coastline, then it should not be rendered as border. This is how also paper maps are usually done. Can you add ticket to the mapnik trac, I think layout bugs are handled through it ? >> If >> Narva city municipal border (and their jurisdiction) ends with riverbank, >> and there is some "federal land" before state/EU border then this should be >> like that in the map also. > > Question: Is this the case or is it just some assumption? Well, just my assumption. Actually http://vana.narvaplan.ee/Linnakaart.pdf shows that the city border is in the middle of the river (like you suggest), but this could also be subjective view of the municipality. Or Maa-amet files are wrong. By the way, http://www.narvaplan.ee/index.php?lang=et&menu=5&page=0 has also Vector map of Narva (in terrible DGN and DWG formats, i.e. probably without attributes, but with buildings). Import? >> Similar thing with Tallinn and Aegna island >> (which is part of Tallinn, Central linnaosa); Naissaar is part of Viimsi >> vald, so it is quite a mixture there if you try to divide sea between them >> reasonably. I've seen that on some maps Võrtsjärv has been divided between >> counties, but in Maaamet latest data has admin borders on shoreline. > > Actually about every map I have seen contains the border(s) in > Võrtsjärv. Yes, it is common on paper maps. It does not mean it is really correct. Maaamet's http://xgis.maaamet.ee/xGIS/XGis does not have any admin boders on Võrtsjärv, Peipsi, Sea and Narva River. It also does not look like a cartographic accident, as on Emajõgi and smaller lakes there are admin borders shown. Either: Maaamet knows that this is right or they just avoid the issue of vaguely defined (undefined) border areas. I wrote emails to Narva city and Võrtsjärve sihtasutus, lets see if they answer anything. ps. We have the final expert questions from KYSK for our project, and deadline until 11.02 to answer them. Jaak _______________________________________________ Talk-ee mailing list Talk-ee@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ee