What ever happened to this project? Is it on ice?

@kaldari I think most of the efforts are currently going towards comment separation/normalization on T6715/T166733 (deployment in progress) and on user normalization on T167246. This is such a generic, epic ticket that most work is done on those subtickets. It is not fast because it requires a lot of code changes, technical debt cleanup (old data in a bad format) and long running schema changes- but as far as I can see it is going well.

There is also work on MCR T174043, which should split the revision table into 2.

I think one those 3 refactorings are done, I assume reevaluation will be done and we will see how compact revision is, improving at the same time how fast newer schema changes can be done.

More specifically,

• T166733: Deploy refactored comment storage will blank the rev_comment field and eventually allow it to be dropped in favor of a bigint rev_comment_id
• T167246: Refactor "user" & "user_text" fields into "actor" reference table will blank the rev_user_text field and eventually allow it and rev_user to be dropped in favor of a bigint rev_actor.
• MCR will allow rev_text_id, rev_content_model, and rev_content_format to be dropped.

In T161671#3889954, @Anomie wrote:

More specifically,

• T166733: Deploy refactored comment storage will blank the rev_comment field and eventually allow it to be dropped in favor of a bigint rev_comment_id
• T167246: Refactor "user" & "user_text" fields into "actor" reference table will blank the rev_user_text field and eventually allow it and rev_user to be dropped in favor of a bigint rev_actor.
• MCR will allow rev_text_id, rev_content_model, and rev_content_format to be dropped.

(Added that last one as T184615.)

After those three components are complete, the idea is to declare this task "done", and if there is scope for further discussion at that point it should be its own task?

Change 350097 abandoned by Brion VIBBER:
WIP - provisional revision table restructure

Reason:
Abandoning this old experimental patch.

https://gerrit.wikimedia.org/r/350097

Changes that affect the revision table will usually also affect the archive table as well. But there is also something that we could do with just the archive table. Perhaps, we could create a page_archive table with columns named pa_id, pa_namespace, pa_title, pa_page_id, and pa_rev_count. Then, the ar_namespace, ar_title, and ar_page_id fields would be migrated to the new table, and we could then add in an ar_pa_id column to the archive table that points to a row in the page_archive table. Also, when a page is being deleted, all the deleted revisions would have a single pa_id, and the total number of revisions the page had prior to the deletion would then become the pa_rev_count. Finally, undeletion would delete row(s) from page_archive table or lower the pa_rev_count field(s) if necessary.

Change 552339 had a related patch set uploaded (by Anomie; owner: Anomie):
[mediawiki/core@master] Alter revision for actor, comment, and MCR

https://gerrit.wikimedia.org/r/552339

Status: This task is now unblocked! I've uploaded a patch for the database part of it, however there should probably be one before it that drops $wgMultiContentRevisionSchemaMigrationStage, or at least makes MediaWiki throw early if it's set to anything other than SCHEMA_COMPAT_NEW.

Recovered!