Re: [Talk-ca] Fwd: [OSM-talk] Argh, Canvec imports
From: James Ewen [mailto:ve6...@gmail.com] Subject: Re: [Talk-ca] Fwd: [OSM-talk] Argh, Canvec imports On Thu, Nov 17, 2011 at 5:18 AM, Richard Weait rich...@weait.com wrote: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: From what I hear, it's people changing the tagging in an effort to tag for the renderer. They are changing tags trying to get the OSM map to look right to them. In doing so, it breaks things. With the bi-monthly rendering runs, it is possible for people to screw up a lot of coastline before finding out that they are breaking it instead. I know that two people on IRC spent a few hours clearing up Hudson Bay and the problems weren't tagging for the renderer, it was mis-tagging and coastlines as multipolygons, of all things. My view is that anyone who is importing in a coastal area needs to have the skillset to make sure that the coastlines are correct, and if they cannot be confident they are not messing the coastline up, they should not import. Personally, I've found the best way to deal with CanVec coastlines is to strip all the tags off of them, merge them to the right layer, and then put the tags on and make sure they're in the right direction. ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [Talk-ca] Fwd: [OSM-talk] Argh, Canvec imports
Hi I /think/ I'm responsible for the problems in Hudson Bay. The issue was I forgot to delete the intermittent coastline from Canvec and only discovered my error once the coastline was updated. I've fixed the problem, but I have no idea if I am responsible for that big blob of grey in the southern half (I didn't do any editing there, I have to wait until the next update to find out). I've seen coastline errors many times before, I remember there was a large one in Southern Ontario for a while, and Newfoundland had several until recently. Here is the problem: -Editors don't notice their mistakes until the coastline is updated, so they stick around in the data for a while -When the mistakes are fixed it takes a few weeks for the corrections to show up on mapnik -Coastlines are handled differently than the natural=water tag, which confuses people -Mapnik doesn't render natural=water at high levels -Mapnik ignores the water=intermittent tag I'm not suggesting we use natural=water (that would be a terrible idea), there just needs to be a way for editors to check the rendering of their coastlines. Sam Dyck ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
Re: [OSM-talk] Argh, Canvec imports
On Thu, Nov 17, 2011 at 08:40, Michal Migurski m...@stamen.com wrote: A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. I care about the coastline too, long time ago there used to be this: http://www.openstreetmap.nl/coastlines.html where anyone could go and fix coastline errors that did show up on a purpose made mapnik layer. I see there is a geofabrik layer for that same purpos, but it is limited just to europe. Michal, are you aware of any such layer, but worldwide? -- -S ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
Quoting Michal Migurski m...@stamen.com: A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk Hi Michal, The rendering error northeast of Quebec City was caused by a novice user who has adjusted the Canvec data which was already there. The import in this region was done by me about half a year ago. The user in question had adjusted the coastline tags to waterway=riverbank, in order for him to show it up properly on his GPS (while using mkgmap for the conversion). He already became aware of what he has done, and has fixed the error himself (which I think is laudable). I've done a check, and the coastline looks all right to me. I'm not sure what to do with the remaining waterway=riverbank tag. This is part of a larger discussion about what to do with large waterbodies. The slippy map doesn't show the corrected situation yet, because the coastline shapefile is not updated very often. Regards, Frank ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
On Nov 17, 2011, at 6:46 AM, Frank Steggink wrote: Quoting Michal Migurski m...@stamen.com: A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Hi Michal, The rendering error northeast of Quebec City was caused by a novice user who has adjusted the Canvec data which was already there. The import in this region was done by me about half a year ago. The user in question had adjusted the coastline tags to waterway=riverbank, in order for him to show it up properly on his GPS (while using mkgmap for the conversion). He already became aware of what he has done, and has fixed the error himself (which I think is laudable). I've done a check, and the coastline looks all right to me. I'm not sure what to do with the remaining waterway=riverbank tag. This is part of a larger discussion about what to do with large waterbodies. The slippy map doesn't show the corrected situation yet, because the coastline shapefile is not updated very often. I had noticed a lot of waterway=riverbank up there, you're right. The first black and white image I link above is derived from the current, November 10 planet dump, so it's a very recent picture of coastline errors. Many of those errors didn't exist in the coastline I made from the November 3 planet dump. Here are some sample locations in Canada where the checker has found errors: http://www.openstreetmap.org/?lat=58.790035lon=-94.202604zoom=17layers=M http://www.openstreetmap.org/?lat=56.749700lon=-92.698650zoom=17layers=M http://www.openstreetmap.org/?lat=45.260826lon=-66.086208zoom=17layers=M http://www.openstreetmap.org/?lat=45.298267lon=-66.117219zoom=17layers=M http://www.openstreetmap.org/?lat=56.970073lon=-92.627314zoom=17layers=M http://www.openstreetmap.org/?lat=47.086208lon=-70.745529zoom=17layers=M http://www.openstreetmap.org/?lat=57.053591lon=-92.00zoom=17layers=M http://www.openstreetmap.org/?lat=45.238757lon=-66.055341zoom=17layers=M http://www.openstreetmap.org/?lat=44.936914lon=-74.704294zoom=17layers=M -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
On 17-11-2011 18:13, David Groom wrote: - Original Message - From: Michal Migurski m...@stamen.com To: Talk Openstreetmap talk@openstreetmap.org Sent: Thursday, November 17, 2011 7:40 AM Subject: [OSM-talk] Argh, Canvec imports A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. -mike. I know, this just looks so wrong to me: http://www.openstreetmap.org/browse/relation/1619985 http://www.openstreetmap.org/browse/relation/1613190 http://www.openstreetmap.org/browse/relation/1618562 and although these relations are not in themselves creating problems, there's an error at http://www.openstreetmap.org/?lat=45.11737lon=-66.52564zoom=15 , which is difficult to fix without knowing how the ways making up these relations are eventually going to end up. Although over the last few months I've been going round trying to fix coastline errors in the rest of the world, I'm afraid that I've virtually given up trying to do anything in Canada David ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk Hi David, Did you already contact that user? This error is caused because the prepared Canvec OSM files also contain waterbodies (natural=water) for areas which are actually in the ocean. Someone importing Canvec is supposed to convert the boundaries to coastlines (except for the sheet boundaries), and replace them with the existing coastline, joining up at the sheet boundaries. Frank ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
On 17-11-2011 16:54, Michal Migurski wrote: On Nov 17, 2011, at 6:46 AM, Frank Steggink wrote: Quoting Michal Migurskim...@stamen.com: A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Hi Michal, The rendering error northeast of Quebec City was caused by a novice user who has adjusted the Canvec data which was already there. The import in this region was done by me about half a year ago. The user in question had adjusted the coastline tags to waterway=riverbank, in order for him to show it up properly on his GPS (while using mkgmap for the conversion). He already became aware of what he has done, and has fixed the error himself (which I think is laudable). I've done a check, and the coastline looks all right to me. I'm not sure what to do with the remaining waterway=riverbank tag. This is part of a larger discussion about what to do with large waterbodies. The slippy map doesn't show the corrected situation yet, because the coastline shapefile is not updated very often. I had noticed a lot of waterway=riverbank up there, you're right. The first black and white image I link above is derived from the current, November 10 planet dump, so it's a very recent picture of coastline errors. Many of those errors didn't exist in the coastline I made from the November 3 planet dump. Here are some sample locations in Canada where the checker has found errors: http://www.openstreetmap.org/?lat=58.790035lon=-94.202604zoom=17layers=M http://www.openstreetmap.org/?lat=56.749700lon=-92.698650zoom=17layers=M http://www.openstreetmap.org/?lat=45.260826lon=-66.086208zoom=17layers=M http://www.openstreetmap.org/?lat=45.298267lon=-66.117219zoom=17layers=M http://www.openstreetmap.org/?lat=56.970073lon=-92.627314zoom=17layers=M http://www.openstreetmap.org/?lat=47.086208lon=-70.745529zoom=17layers=M http://www.openstreetmap.org/?lat=57.053591lon=-92.00zoom=17layers=M http://www.openstreetmap.org/?lat=45.238757lon=-66.055341zoom=17layers=M http://www.openstreetmap.org/?lat=44.936914lon=-74.704294zoom=17layers=M -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk I've fixed some coastline issues at the Nelson River / Hayes River mouths in the Hudson Bay, see [1]. There were some duplicated coastlines, and an inland island (which was running the wrong way). Near Churchill, MB [2] the coastline was already fixed by the Canvec import user a few days ago. I didn't see any coastline mistakes there. However, the treatment of intermittent water is far from perfect, and not how I would have done it. I'd rather convert it to natural=wetland, wetland = tidalflat, indicating that mudflats show up at low tide. Doing this along the entire Canadian coastline, which is very extensive, is a tough task. The coastline is really peculiar in some places, like here: [3], so I'm afraid we haven't seen the last error. Tools assisting us in error checks would be great, and absolutely necessary in the Canadian case. Frank [1] http://www.openstreetmap.org/browse/changeset/9856941 [2] http://osm.org/go/W6dBGeK- [3] http://osm.org/go/W7CeaB ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
- Original Message - From: Frank Steggink stegg...@steggink.org To: talk@openstreetmap.org Sent: Thursday, November 17, 2011 7:03 PM Subject: Re: [OSM-talk] Argh, Canvec imports On 17-11-2011 18:13, David Groom wrote: - Original Message - From: Michal Migurski m...@stamen.com To: Talk Openstreetmap talk@openstreetmap.org Sent: Thursday, November 17, 2011 7:40 AM Subject: [OSM-talk] Argh, Canvec imports A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. -mike. I know, this just looks so wrong to me: http://www.openstreetmap.org/browse/relation/1619985 http://www.openstreetmap.org/browse/relation/1613190 http://www.openstreetmap.org/browse/relation/1618562 and although these relations are not in themselves creating problems, there's an error at http://www.openstreetmap.org/?lat=45.11737lon=-66.52564zoom=15 , which is difficult to fix without knowing how the ways making up these relations are eventually going to end up. Although over the last few months I've been going round trying to fix coastline errors in the rest of the world, I'm afraid that I've virtually given up trying to do anything in Canada David ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk Hi David, Did you already contact that user? Frank I haven't contacted the user, because from reading http://wiki.openstreetmap.org/wiki/Canvec I could not find enough information to tell him what he should be doing. I will however now contact him and copy the information you have given below, and suggest he ask on talk-ca if he has any other problems Regards David This error is caused because the prepared Canvec OSM files also contain waterbodies (natural=water) for areas which are actually in the ocean. Someone importing Canvec is supposed to convert the boundaries to coastlines (except for the sheet boundaries), and replace them with the existing coastline, joining up at the sheet boundaries. Frank ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Argh, Canvec imports
On Nov 17, 2011, at 12:24 PM, David Groom wrote: Although over the last few months I've been going round trying to fix coastline errors in the rest of the world, I'm afraid that I've virtually given up trying to do anything in Canada David Hi David, Did you already contact that user? Frank I haven't contacted the user, because from reading http://wiki.openstreetmap.org/wiki/Canvec I could not find enough information to tell him what he should be doing. I will however now contact him and copy the information you have given below, and suggest he ask on talk-ca if he has any other problems Thanks David, you've got more patience for this than I do. I forced myself to learn JOSM to fix some of the river/coast interactions around Montréal. =) -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk
[Talk-ca] Fwd: [OSM-talk] Argh, Canvec imports
see the conversation around changing coastal tags on talk@ -- Forwarded message -- From: Michal Migurski m...@stamen.com Date: Thu, Nov 17, 2011 at 2:40 AM Subject: [OSM-talk] Argh, Canvec imports To: Talk Openstreetmap t...@openstreetmap.org A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list t...@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk ___ Talk-ca mailing list Talk-ca@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-ca
[OSM-talk] Argh, Canvec imports
A cry of frustration: I do bimonthly runs of the worldwide coastlines (http://metro.teczno.com/#coastline), and Canada seems to be a recurring source of problems. What the heck is going on up there? I consistently see new imports of what seems to be Canvec data screwing up coastlines and making for some deeply broken renders: http://mike.teczno.com/img/broken-coast.png Most of that junk in the Atlantic Ocean is newly introduced within the last few weeks, and is making it difficult to get out a clean coastline suitable for rendering. It shows up in the main OSM mapnik tiles in Hudson Bay, too: http://www.openstreetmap.org/?lat=55.7lon=-83.6zoom=5layers=M I've been trying to fix problems as I encounter them and I've managed to fix invalid coastilnes around Baltimore, Houston, Tampa and the dreaded Montréal area in an effort to generate a usable map of North America, but this Canvec stuff is absolutely killing me. -mike. michal migurski- m...@stamen.com 415.558.1610 ___ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk