Topic on Talk:Multi-Content Revisions/Database Schema

Summary by Daniel Kinzler (WMDE)

The discussion on wikitech-l indicates that we are not ready to drop rev_sha1 at the moment. https://lists.gt.net/wiki/wikitech/844067

We could however re-calculate rev_sha1 for every revision based on the actual blobs, instead of using pre-calculated content_sha1 fields. We would trade the overhead of loading and hashing all content against the overhead of storing the pre-calculated hahes in the DB.

Anomie (talkcontribs)

That is, rev_sha1, ar_rev_sha1, and content_sha1. Other teams have been discussing revert detection, which is the only point of having these fields. It would be useful to get their input on these. Post to wikitech-l?

Daniel Kinzler (WMDE) (talkcontribs)

I was thinking of turning the schema proposal into an RFC anyway. But I can send a separate email right away, sure.

I believe the person looking into revert detection/tracking is Roan.