• Read through Phabricator and code bases for TODOs and previous attempts to rationalize sites.
• Read code for third-party wikifarm integrations, eg. Miraheze
  • Miraheze has a custom extension https://github.com/miraheze/WikiDiscover as its source of truth. This manages a database table cw_wikis.
  • They are running https://github.com/miraheze/MirahezeMagic/blob/main/maintenance/PopulateWikibaseSitesTable.php but would be happy to drop if unnecessary.
  • Current working theory is that the WikiDiscover extension can provide a custom SiteLookup and configure that as the default sites service.
• Read MediaWiki-core sites code
• Read mediawiki-config wiki farm setup and scripts
  • Source of truth are the dblist files which are modified using composer manage-dblist
• Read ext-Interwiki code
• Read ext-SiteMatrix code
  • The source of truth comes from several places: $wgSiteMatrixSites defines the sister projects; languages come from mediawiki-config langlist, and whether or not a language-project (eg. "eswikibooks") actually exists is determined by whether it is listed in all.dblist which gets loaded into $wgLocalDatabases.
  • Working theory is that SiteMatrix can provide the default SiteLookup, when the extension is enabled.
• Read ext-WikimediaMaintenance scripts

https://docs.google.com/spreadsheets/d/16v1kddQS0Kr--zdyRv3XyiYk2g9xjc_3wED1eePUlnQ/edit?gid=0#gid=0