It is a generic problem, but it has particularly nasty interactions with
RevisionDelete: fundamentally, it can become impossible to tell when, why or
by whom a RevDel'd revision was deleted. The other manifestations of the
lunacy of identifying deleted revisions by timestamp rather than revision id
are not so serious, but these definitely are.
Identifying deleted revisions by id rather than timestamp is a lot of work
in itself; not least of which is writing and running a script to fill in all
the null values for old deleted revisions on WMF wikis. There's no point in
doing that if the 'grand plan' is to move away from having an archive table
altogether. But equally this is an issue which does badly need work done.
Hence the question, ""what *is* the grand plan?""
--HM
--------------------------------------------------
From: "Andrew Garrett" <andrew(a)werdn.us>
Sent: Tuesday, May 18, 2010 3:05 PM
To: "Happy-melon" <happy-melon(a)live.com>om>; "Wikimedia developers"
<wikitech-l(a)lists.wikimedia.org>
Subject: Re: [Wikitech-l] Deletion schema
On Tue, May 18, 2010 at 8:24 AM, Happy-melon
<happy-melon(a)live.com> wrote:
There's another discussion happening at
enwiki at the moment about the
stalled rollout of RevisionDelete for admins; which is backed up in the
chain of bugs which boils down to "our deletion mechanism is borked".
Reviewing the whole deletion mechanism was on the topic list for the last
dev meetup, but AFAIK despite that event running for three times as long
as
it was expected to, it never got raised? I think this would be as good a
time as any to do so. Do we have any clear idea or overall plan for page
and revision deletion, the archive table, a page_deleted field, a
deleted_page table, etc etc??
I promised to activate single-revision deletion for admins months ago,
and then again after the last software update. I finally got to it
tonight.
Church of Emacs points out this bug, but it looked as though it was a
generic problem with our previous deletion system, and not an
additional problem caused by single-revision deletion (indeed, it
appears to be mitigated by the use of single-revision deletion instead
of the old delete/undelete method).
https://bugzilla.wikimedia.org/show_bug.cgi?id=21279
--
Andrew Garrett
http://werdn.us/