4 ms·
Some people would take issue with the "It works well" part. https://lists.wikimedia.org/pipermail/wikidata/2020-February/013793.html https://lists.wikimedia.org
by zozbot234 7y ago
Some people would take issue with the "It works well" part. https://lists.wikimedia.org/pipermail/wikidata/2020-February/013793.html https://lists.wikimedia.org/pipermail/wikidata/2020-February... Turns out that scaling up a general-purpose triple store beyond 10B triples is really, really hard, and that few or no people in the FLOSS community (important due to Wiki's reproducibility and freedom-to-fork requirements) are paying attention to the problem.
- thom 7y agoThis is a shame, because I genuinely find SPARQL over Wikidata one of the most insanely magical things on the internet. It is literally like being able to just query the universe directly and there's almost no end of fun that can be had formulating new questions you can answer and then finding out it's actually possible. Of course I now feel bad for occasionally hitting it with my frivolous queries, and would be very sad to see it limited.
- ansible 7y agoMediaWiki (the actual Wikipedia website) runs on top of a traditional SQL database (originally just MySQL, but others are now supported too): https://www.mediawiki.org/wiki/Manual:Installation_guide https://www.mediawiki.org/wiki/Manual:Installation_guide There have been various forays into Semantic Web technologies. Previously people have done things like parse the infoboxes on Wikipedia pages and then import that into a triple store. It seems that now the Wikimedia Foundation is finally interested in moving in this direction. But the software isn't there yet, certainly not at the scale Wikipedia needs.
- zozbot234 7y agoThe website you link to states that the support for non-MySQL-derived db's (PostgreSQL and SQLite) is buggy and that not many people are working on it. It seems that legacy constraints are essentially forcing them to stay with their old flaky PHP codebase and MariaDB database. So much for that blockchain thing!
- spauka 7y agoWhere does it say that? My reading of the linked page suggests that MSSQL and Oracle are being deprecated, but SQLite and Postgres are supported and maintained. In fact, SQLite is the first database listed on the detailed install instructions page[1]! [1] https://www.mediawiki.org/wiki/Manual:Installing_MediaWiki#Create_a_database https://www.mediawiki.org/wiki/Manual:Installing_MediaWiki#C...
- ansible 7y agoI've been running a small, low-traffic SQLite3 instance for many years. Actually, too many... I need to go and upgrade it.
- jonathanpierre 7y agoIt does work well though. As a user, I can't remember significant Wikipedia or Wikidata outages. And I'm really not sure if whatever blockchain solution that person on twitter wants Wikimedia to use is even able to get to that scale, yet alone perform better an it than what Wikimedia currently has.
- diegoperini 7y agoHere in Turkey, we had Wikipedia banned for 4 years. Luckily, the ban has been lifted last month. I'm not arguing blockchain would be a good mitigation for a future incident though.
- JshWright 7y agoWas that the fault of Wikipedia's database?
- samatman 7y agoIn a limited sense, yes, it was. Because Wikipedia's database is centralized, and lives at a particular (range of) IP addresses, it's trivial for a state actor to censor it. Does a sprinkling of magic blockchain solve this problem? Not really. But something more specific to the problem might; being able to torrent the full data dump is a step in the right direction.
- JshWright 7y agoUsers do not connect directly to the database. Turkey definitely didn't blackhole traffic to Wikipedia's database.
- icebraining 7y agoThe fact that users can't connect directly to the database is, at least in part, a limitation of the database chosen.