Topic on User talk:Anomie/Abstract schema

If you have a wiki in postgres, there's no good migration path to mysql. Could we introduce some abstract SQL dump format? Could also be used by Wikimedia as well (in addition to the mysql dumps).

It would be neat if the schema definition could indicate whether the data in the column is private or not, so that the abstract dump could just omit that information if --public or something was provided to the dump script (T140977). It could also (partially?) replace the logic that's maintained in maintain-views.yaml.

Probably a separate project in itself, but wanted to mention the idea to make sure any necessary groundwork wouldn't be missed.

The abstraction layer would have to be revdelete-aware to properly support --public.

An "abstract dump format" seems outside the scope of this proposal. Although this proposal's implicit fixing of various schema inconsistencies might make it more feasible to use pg_dump --inserts or other existing tools to do such a migration.

Indicating public data seems far more complex than something sensibly maintained in the core schema, as is indicated by the complexity of the logic for some tables in the file you link.

I think a private/public flag, or the ability to flag and annotate specific fields, would be useful functionality. Even if it were like a comment/docstring near the schema definition for a particular table column?