On Mon, Jun 12, 2017 at 1:32 PM, Roland Olbricht <roland.olbri...@gmx.de> wrote: > Hello everybody, > > first of all, I would like to thank Michał for the analysis. However, after > cross-checking three random changesets, I come to a different conclusion. > (...) > Thus, in total we got meaningful observations from surveyors that were > there. Hence, it would be outright vandalism to revert them.
Well, I have to admit leaning too much on the formal side (however funny that sounds in a folksonomy tagging system :P). However, I don't advocate reverting maps.me edits on sight if they simply contain errors. I don't think Tomas meant it either. See how 8% was garbage? For a big country that can easily be a few per day. I've seen people add embassies in absurd places (because "Ambasada" is first alphabetically). The issue that I wanted to point out is that many of these errors would be preventable if MAPS.ME did things right and did due diligence. Regarding Tomas' frustration, I can somehow relate. MAPS.ME-only mappers almost never reply to any changeset comments, however nice you talk to them. Even though MAPS.ME has been contacted about it, the issue remains unresolved This is so frustrating to many of us, because communication is a pillar of efficient community. iD beats it by a quite a margin, there got to be a dozen or more of iD noobs that besides responding to my comments even apologized for their mistakes. > As often told: OpenStreetMap values local contributions over mechanical data > conversions, regardless whether these are imports or linting. You can have > an opposite mind, we do not harass you on that. Please take a copy of the > database and do whatever processing you like on that database. > > As this has been told again and again, I would like to explain it this time > a more explicit way: > > The price to reproduce all the on-the-ground mappers' contributions we have > is likely to be somewhere between 4 million and 150 million euros: to > reliably tell where all the pharmacies and all the buldings are, and what > names the streets have, we would have to traverse all the roads in the > world. > > (...) > > Hence, walking-and-mapping the world would cost somewhere between 4 million > and 150 million euros. > > By contrast, to find and fix all places with "name" equals "name:pl" in a > neighbourhood is just a click away: > https://overpass-turbo.eu/s/pGL > A competent freelancer would build for your within a day or less a script to > drop all such tags from your copy of the database, at 1000 to 2000 euros. > Keeping up a server with that scrubbed data may cost another 500 euros per > year. Even if you plan to keep your copy online for 20 years then it has a > total value of at most 12000 euros. > (...) > So, dear bot editors, you are jeopardizing 150 million euros to save some > thousand euros at best. This is why we, the community, ask you to be > extremely cautious or not doing it at all. I don't quite get it. The monetary argument is quite far-fetched and non-sequitur. The fact that idiots did and do bad imports / mechanical edits doesn't automatically validate just _any_ edit made on local knowledge. With knowledge, there should also come skill, which MAPS.ME editors won't get because they don't respond to changeset comments! Don't get me wrong, I would be happy to teach someone how to become a good mapper, even if that would require a level of patience, but because of no contact, I can't. The problem with name:pl and the likes is akin to DB denormalization. Should a name change, I bet there's quite a non-zero chance someone will desync name from name:pl Similarly, with addresses I can see geocoders freak out on malformed ones. On a side note, as I and Zbigniew_Czernik say, "bad data introduces more bad data". People learn by imitation. That's why we should hold data to at least a modest standard. The takeaway is that many of these issues would be preventable if MAPS.ME only devoted resources to fix them. But case after case, they don't see a problem. Based on my analysis and general experience with fixing MAPS.ME edits, I would recommend this: * Discourage entering name:xx if xx is spoken in that country. * Add support for addr:place (display and editing). * Ensure that we have addr:housenumber and (addr:street or addr:place) and that addr:postcode can only be entered on such correct addresses. * Explain to users that they should not add things if they don't fit any of the available categories. * Add more POI categories so that wouldn't be the case so often. * Try at least heuristically detect duplicate POI to be added and ask the user if it's not a mistake. * Notify users of changeset comments and allow discussion in-app using OSM API. * Be more strict about using old versions to edit, like StreetComplete blacklist (a mere warning would be OK actually). * Try to cooperate with the community so that *all* MAPS.ME edits are at least formally (armchair) verified - by extending mmwatch.osmz.ru with means of marking changesets "verified"/"unsure"/"bad". * Reword "add a place to map" so that it's evident it goes into OSM. Or remind the user where personal bookmarks are. That alone would cut down BS edits and notes significantly. In general, the process of sending edits isn't thoughtfully designed. Now, it works like that: some changes are lumped together and then a changeset is uploaded in the background. If I were to design it, I would make some sort of upload prompt, with user name and avatar in the header (so they know it's public) and a list of changes to verify once again. I understand that the current upload logic makes it convenient. But there is a product design rule that people value things more if they have to put some little work in it (the IKEA effect). By carefully laying out and wording such upload dialog, they could eliminate many of sloppy and garbage edits, all while increasing retention. People who edit with MAPS.ME should be well aware that we are not some paid-for moderators, but a volunteer community of people just like them and MAPS.ME is only one way of contributing to OSM. IIRC, if you go into your profile in MAPS.ME it says "changes verified: <number>" which is very misleading. On the interface between MAPS.ME and the community, there should be some dedicated person to discuss issues while being an insider to the project (ie. not some PR drone). The current "we don't look at GitHub issues, please send them to b...@maps.me where nobody outside of our team will see it" situation is clearly not in our interest. Remember that people aren't angry at MAPS.ME "just because", there is concrete evidence of low edit quality associated with this editor, which haven't been addressed. The situation wouldn't look like this if effort was put to fix them. We can cooperate, but that needs addressing the issues on their side. Michał _______________________________________________ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
