On 08/01/14 16:57, Denny Vrandecic wrote:
I am afraid that keeping versions is something the current architecture will be terrible at supporting, especially because we need to keep snapshots over a multitude of pages. MediaWiki is not terribly good at that. If something like the memento extension would really work at scale, maybe...
One could possibly keep versions of the data (only) in a secondary archive that does not need to feed back into Wikidata. However, the question is whether this would help for the problems that Thomas mentioned: if someone is frustrated because a property was deleted unexpectedly, then it will not be enough to provide access to old data that still used the property.
One other idea could be to support more types of "views" in top of current data. If some property is deleted, it is often replaced by more appropriate information or by some new encoding. If there is a systematic way to compute the value that the old property should have had from the new data, then one could provide access to this data as if it was still in the system. Again this would be done in external systems, so as to not make Wikidata any heavier. It would be good to have concrete examples of problematic schema changes to see what is needed to build such a view (and if it can work in the problematic cases).
Cheers,
Markus
Anyway, I guess going with the watchlists sounds powerful. Notifications would require checking the wiki (right? I am not completely sure about that) whereas watchlists provide feeds, which could be integrated into a large system.
Just my two cents.
On Wednesday, January 8, 2014 7:21:10 AM, Thomas Douillard <thomas.douillard@gmail.com mailto:thomas.douillard@gmail.com> wrote:
Hi, a problem seems (not very surprisingly) to emerge into Wikidata : the managing of the evolution of how we do things on Wikidata. Properties are deleted, which made some consumer of the datas sometimes a little frustrated they are not informed of that and could not take part of the discussion. My question is : is it a community problem, a technical problem, or both ? IMO it is a very serious problem for a project of the size of Wikidata that just leaving to humans to make notifications to whoever uses te datas could be a disaster. We therefore need to have tools to manage that, part of the solution is from the technical side. I'll try to do a review of which tools we have now, as tools, to make notifications from people who want to make a change that imacts other projects or people, and interestec projects or people: * Purely human : someone who make a change in a model or delete a property is responsible to notify every projects he knows would be impacted. This is tedious and imperfect, and imperfectly pass through language barriers, if he has to notify every Wikipedias. * Using watchlists : This is a semi automated process. Every people involved into a project, model, or property has to follow the relevant pages. Still imperfect for many reasons. * Using the notify extensions, like the ping projects or ping user templates, interproject, this may need software solutions from the developper of the Echo Mediawiki extensions Another, and complemnetary solution is to mantain data or models API and versions, as we do in sotfware API, and manage versioning, parralel versions, deprecations ... but this is ressource consuming. Any ideas ? I thing it's an important question we need to give answers, at least partial ones.
Wikidata-l mailing list Wikidata-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikidata-l