On Fri, Apr 3, 2009 at 12:01 PM, Robert Rohde <raro...@gmail.com> wrote: > That said, I'm also not sure why we couldn't trust sysops with this. > Assuming the limitation is not technical (i.e. the servers won't > explode)
I think Domas might have objections if the number of renames suddenly increased by a factor of 10, actually. He's complained about them before. They have to update an awful lot of rows. > and renames are undoable (they are right?) In principle, yes. Of course they can get stuck half-done, and if you rename someone twice in quick succession I don't know what will happen with the job queue. Renaming is not currently a simple operation, because of our schema. We encode the username all over the place. If it were just in the user table, renaming would be less of a big deal, since we'd change one row plus maybe some stuff in RC. But we have to change a ton of rows in a ton of tables, almost all of which (for long-established users) will be cold and so will be dug out from disk. So we have to defer things, so they can be in a half-finished state, so it's slow and messy. Also keep in mind that if someone is renamed without their knowledge, they'll be unable to log into their account. The error message will only say that the user doesn't exist, or -- if a user is created with the same name -- it will just tell them their password is wrong! Forcible renames are scary for this reason as well. _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
