I am proposing to remove all brand:wikipedia tags from all POIs in the United Kingdom.
I don’t think I can put it better than the successful proposal to do the same in the US a couple of years ago:
Problem Statement
brand:wikipedia is no longer automatically maintained by iD and thus is likely to become out of date over time. In the words of @Minh_Nguyen:
“if there’s any mismatch between the brand:wikipedia and brand:wikidata tags, brand:wikipedia is almost guaranteed to be wrong.”
brand:wikipedia provides no additional data not already provided for in brand and brand:wikidata, which are maintained in Name Suggestion Index and updated automatically by iD. As iD updates the other two brand tags, brand:wikipedia remains untouched. The brand tag provides a human-readable brand name, so there is no need to keep brand:wikipedia around merely for human readability.
The existence of this tag on current objects creates an implied need for mappers to maintain them, when such maintenance is useless work when it can just be removed and mappers can focus on more important work.
Removing this tag removes the risk that an unaware data consumer will consume the brand:wikipedia tag and ingest wrong information from OSM.
The only thing I would add is that of course the Name Suggestion Index is used by more than just iD, at least a number of mobile editors and is also available in JOSM.
The tag has been seeing a decline in usage since 2022 (with a blip where there was an undiscussed attempt to remove it, which was then reverted).
An argument against doing this would be that it updates the most recent edit date of the elements concerned. Therefore, a possible option would be to add check_date for the most recent edit date of each element concerned to mitigate this issue (if check_date is not already tagged). (I know that a ‘better’ option would be to analyse when each element was last actually surveyed and add the check date for then, but with over 70000 objects that would not be feasible.)
I have not yet written any code that could add the check_date, but I am sure that I would be able to, if that were the preferred option. If this were approved I would do that and then perform the edit in a series of changesets geographically limited in scope.
Please vote for your preferred option and comment on any issues that you see.
Remove brand:wikipedia
Remove brand:wikipedia and add a check_date of the most recent edit
Fully agree that these should be removed, wikipedia article titles aren’t stable identifiers, and it can even require a tag for each language eg. brand:wikipedia:en=*
The whole point of wikidata is to provide this stable identifier that is easy to lookup and links to other languages etc.
Maybe an unpopular opinion but even the wikipedia tag itself is mostly redundant compared to having a wikidata tag that can properly link to thousands of other databases and articles in other languages, while being a stable identifier.
I don’t agree that it doesn’t matter much, if the article title changes, the data in OSM is now wrong and requires time to correct that could be spent doing something actually useful, or someone has to maintain a bot (which is unlikely to happen). There’s no reason to subject the project to this when there’s a perfectly good stable identifier available.
Last edit date is already fubar in most cases, moving it to check_date wont help. eg the only real check_date for this is 2009-12-06.
I would like it if metadata like last edit date meant something and were respected. That ship might have sailed, but I don’t want check_date to be ruined too.
My vote would be to normalise removing brand:wikipedia (when brand:wikidata is present). That would involve it being acceptable and acknowledged as redundant/potentially harmful, editors (iD, StreetCompete) doing it automatically and people mapping.
This should only happen where there is a wikidata tag.
Sometimes the name, brand, brand:*, wikidata, and wikipedia do not agree, as someone has changed one or more tags and it hasn’t carried through to the others.
Sometimes, there are subtle differences in identity, e.g. a brand is owned by a foreign conglomerate and there is a difficult-to-define overlap between the foreign conglomerate and local brand that wikidata hasn’t been able to deal with.
To elaborate, besides the usual reasons, these tags can get out of sync because Wikidata is more granular than Wikipedia. For example, the English Wikipedia only has a single article “Shell plc”, formerly titled “Royal Dutch Shell”, while Wikidata has items about the UK company versus the petrol station chain.[1] Usually Wikidatans only get around to creating the more specific items when there’s an immediate need, like the brand’s owner getting bought up or franchising the brand in certain countries. Once that split happens, brand:wikipedia=* goes stale, because the original item usually belongs to the more “notable” company. Sometimes it’s possible to link the new item to a redirect page on Wikipedia, but otherwise, no brand:wikipedia=* can possibly apply.
I think there used to be a separate item about the former Anglo-Dutch firm, but Wikipedians keep unilaterally merging these items incorrectly, unaware that Wikipedia:Notability isn’t the end all and be all. ↩︎
If the stuff is connected with NSI, there should be a key like brand:ref:nsi. The inconsistent Wikidata stuff just creates an additional layer of discrepancies.
With currently 30 votes for and 2 against, and no further arguments against the edit, I am currently treating this as approved. I will give a few more days for any more comments or votes before performing the edits.
Arguably it should be cited in the changeset rather than on individual objects. But as such a suggestion is completely unrelated to the proposal, I would encourage @CjMalone to start a new thread (or split the thread) order to keep this one on track.
