Hi, On 10/06/10 21:28, john whelan wrote: > If we are importing then we can pull in the stop_id for free. Having it > available means an application can then make use of it or a bot can > verify the values in the other fields more easily. In database terms > having a value that is dataset unique is important.
I just wanted to emphasize how important I think that is. When importing stops into OSM, I think it's vital to include: * some value that uniquely identifies the data source, world-wide (in the UK this was tagged as 'source=naptan_import' although I may have preferred 'uknaptan'); I don't think it's necessary to identify the version of the data source; * any unique stop ID that exists in the source data; it shouldn't matter how this gets tagged in OSM, as long as it's done in the same way for all data from a particular dataset. In the UK I think 'naptan:AtcoCode=*' is being used for any data from that source. If there is no nation-wide numbering scheme for stops, make sure to uniquely identify the dataset you're using (eg. the country, county and/or operator it relates to) and import whatever unique ID you can. This is important because, at a later time we'll likely want to merge changes in from new versions of the source data -- and perhaps even merge changes in OSM back to the data source, too. As long as we have those two things imported, we can handle new versions of the dataset like this: * for each item in the new version of the dataset: ** compare with corresponding record in the old version of the dataset ** if different: *** search OSM nodes for 'source=*' tag corresponding with this dataset AND unique ID (eg. 'naptan:AtcoCode=*') matching this record *** if found, AND changed data fields (eg. lat/lon) in OSM still match those in the old dataset: **** update OSM node with the new data (eg. lat/lon) *** otherwise, flag this as a conflict to resolve manually, or as a new stop to import This process could be very problematic to automate if no dataset name or unique IDs were imported into OSM; there would be a *lot* of guesswork (and probably a lot of mistakes) and/or updating by hand otherwise. Regards, -- Steven Chamberlain [email protected] _______________________________________________ Talk-transit mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk-transit
