I welcome this initiative from Ilya. We are never going to get a totally correct data source, and even data originating from human mappers can leave a lot to be desired.So let's get beyond criticising the acuracy of the data and start working out how we're going to tackle this On the whole this import I believe improves our data coverage - I suspect that in the end the accuracy will be prove to be in the region of 80-90% - (everything within a 5 mile radius of me is perfect). . When we imported Naptan data we added a review tag set to "no", so that as the community verified the data it could be set to "yes" or just deleted. That would give us a method of knowing what's accurate and what's not and what's actually been reviewed on the ground. An overpass query could easily measure progress. It would make a great quarterly project!
Regards Brian On 3 November 2017 at 12:06, Andy Townsend <[email protected]> wrote: > On 03/11/2017 10:59, Ilya Zverev wrote: > >> >> How is the address info there incorrect? Did you see the correct address >> of the building? Googling shows that amenities there are indeed addressed >> by Markham Lane, just like the Markham Vale, the business centre that >> contains these. >> > > I explained the likely chain of events the last time this was discussed: > > https://lists.openstreetmap.org/pipermail/imports/2017-May/004985.html > > Basically, when the place was built the road it is now on > http://www.openstreetmap.org/way/393054254 (which based on the signage to > the south and http://ris.dev.openstreetmap.org/oslmusicalchairs/map?osl_id > =1001791 is called Enterprise Way) did not exist, and the site office was > where Google thinks everything still is, in an office off Markham Lane. > > > > I have received the list from the Shell UK. Are you suggesting I should >> buy a ticket and verify each of these on the ground? Are there any better >> sources for Shell fuel stations, or are you implying the UK is a special >> restricted no-imports-whatsoever territory? >> > > No, I'm suggest that you (or whoever's doing this import) sanity-check the > data that you get from third parties against what's already in OSM. David > Woolley's already said "company head offices are notorious, in my view, for > failing to map their branches at the correct location"; here it looked like > someone at Shell had to fill in address details on a form, but couldn't use > e.g. the correct road because it did not exist yet, so made up some values > to fill in the fields to avoid "computer says no" and that's never been > updated at Shell's end. > > > >> I assumed a week of heated discussions in imports@ and talk-gb@ mailing >> lists in May would be enough, and we won't start with the same arguments. >> > > What would be good would be if you fixed the problems raised there. Street > names were raised as an issue here: > > https://lists.openstreetmap.org/pipermail/imports/2017-May/004974.html > > > >> ... I personally believe our map would be better by having more data from >> the original sources: >> > > Agreed, if the data is correct (and it's going to be maintained, as I > presume it is here). The problem here seems to be that it isn't correct in > the first place. > > > >> I see many imports every day going unnoticed, and I see this kind of a >> harsh reaction on proposed imports. It is easy to see why people are >> reluctant to announce their imports and automated edits beforehand. Until >> we develop a polite, predictable and mature way of clearing imports, we >> will continue seeing undiscussed imports and being angry at people who just >> did not want to be yelled at. >> > You're not being yelled at. You're being told that some data that you > want to import is incorrect, as you were back in May. > > Where data may be iffy / need conflation there are are lots of > alternatives to "let's dump the data in OSM and let the community figure it > out after the event" - for example look at the way schools (including in > relatively unmapped areas) were mapped in the UK as part of a project. > Shell stations by there very nature tend to be on major roads; I'd be > surprised if many of the new ones (and things like brand changes) weren't > visible from e.g. Mapillary to at least verify the location. Similarly, > any human that looks at the address example I gave ("house number 29a", > which is obviously the motorway junction ref) is unlikely to be valid. > Similarly for the address issues that Richard has raised. > > Best Regards, > > Andy > > > > _______________________________________________ > Talk-GB mailing list > [email protected] > https://lists.openstreetmap.org/listinfo/talk-gb >
_______________________________________________ Talk-GB mailing list [email protected] https://lists.openstreetmap.org/listinfo/talk-gb
