Finished my own patch (after 2 hours!): http://www.openstreetmap.org/browse/changeset/9093636
Uploaded in ~2000 object chunks prevented "Ignoring caught exception because upload is canceled. Exception is: org.openstreetmap.josm.io.OsmTransferException: java.net.SocketException: Socket closed" There are a few unconnected/untagged nodes outside (south) my patch, will cleanup tomorrow. I'm assuming it is an upload in process. On Mon, Aug 22, 2011 at 9:49 PM, andrzej zaborowski <[email protected]> wrote: > Some notes from me, I'm still working on a tile of the MadMappers data. > > On 22 August 2011 14:51, Harry Wood <[email protected]> wrote: >> Download the chosen .osm file and open in JOSM >> >> File -> new layer then download OSM data covering the same area (and a bit >> more). I needed to do this peice by peice because the are was too large for >> a single download (although not much data being returned by the server) >> >> Using the layers panel to flick between datasets. Visual compare. look for >> overlapping areas for road. > > I prefer using the mapnik tiles (Imagery menu) as a background for > finding the duplicates, makes it easier to see which data is the new > and which is existing OSM data because of the different style. > Obviosuly some features are not on mapnik so it needs and additional > check later. > >> >> For me there were no existing roads in OSM to worry about apart from in the >> town on Burao http://www.openstreetmap.org/?lat=9.5151&lon=45.5441&zoom=14 >> I did a general sanity check on the rural roads data, comparing bing >> (landsat) The MadMappers roads are mostly not visible in this, but they >> seemed follow logical well lined-up paths through mountains and along valley >> bottoms. I decided to upload the data to OSM first (creating duplicates in >> Burao) and then make a few de-duping fixes after that. > > In my case there's lots of existing data so there's lots of work on > deciding which is more exact in a given area and it's probably a good > idea to do the fixing up (deleting ways and joining at crossing) all > in the same changeset. Or maybe that's just a programming habit to > preserve bisectability :) > >> >> Go to openstreetmap.org 'logout' then 'sign up' to create a new user >> account for imports (if you don't already have one) e.g. 'Harry Wood >> Import' Verify email address etc. >> >> back in JOSM Edit preferences. 'Connection Settings' panel. Swap to new >> username and password >> >> Click upload. Set the source=MadMappers-SOM-2008-01 on the changset >> >> I had 13,000 objects. This took 20 or 30 minutes to upload as a diff. I'm >> not sure if there's a better (more robust and less strenuous on the server) >> way to do this. Perhaps switching to "atomic" mode would be a good idea for >> this. > > There's probably no faster/less strenuous way but make sure to use an > atomic upload if you have < 50000 elements, rather than upload in > chunks, else there's always a risk of a broken upload. > >> >> back in JOSM Edit preferences. 'Connection Settings' panel. Swap to my >> normal username and password. >> >> Delete all layers and re-download the OSM and MadMappers data (now munged >> together on the server) >> >> --- Fixup --- >> From this point onwards I was just doing normal editing. Uploading >> regularly, and I didn't set a source tag (what is the source for fixing up >> an import? 'common sense'?) >> >> The existing data by User:roage was a few main roads in the town. These >> lined up pretty well with the data from MadMappers, but was more crude. It >> mentioned source=landsat in various places, although I hated to do this, I >> mostly deleted User:roage's contributions in favour of the MadMappers data. >> However User:roage's data included stuff off the edge of the area I was >> dealing with so this survived. I had to cut and re-connect these ways with >> the imported data (main roads coming into Buroa from the south west) >> >> In the MadMappers data itself there were some duplicates in the town. I >> didn't seed ways directly on top of each-other, but there was highway=track >> alongside highway=secondary. These were side-by-side like dual-carriageway >> or in places braiding over the top of each-other. Although this might have >> reflected some sort of reality, I'm guessing not, particularly as the >> highway=secondary was also surface=unpaved anyway. Clearly there is one main >> (but unpaved) road following this route through the town. So I guess this is >> the "doubled/overlapping roads" Schuyler is referring to. I imagine this is >> probably data from two surveyors munged together messily. So I deleted the >> highyway=track and made the necessary reconnections. > > Additionally I've found a couple of named tracks which seem to > actually represent areas, not routable tracks. I made the shapes > closed (they had small gaps) and tried to derive the tagging from the > name ("governors quarters", "meat packing plant", etc) > > I also decided (wrongly?) to change all highwy=unclassified's to > highway=residential because all of them were in a town and it's > unlikely that there would be any roads that count as > highway=unclassified inside a town, please correct me if I'm wrong. > > Cheers > > _______________________________________________ > HOT mailing list > [email protected] > http://lists.openstreetmap.org/listinfo/hot > -- cheers, maning ------------------------------------------------------ "Freedom is still the most radical idea of all" -N.Branden wiki: http://esambale.wikispaces.com/ blog: http://epsg4253.wordpress.com/ ------------------------------------------------------ _______________________________________________ HOT mailing list [email protected] http://lists.openstreetmap.org/listinfo/hot
