[This is long, but hopefully helpful. You may want to skip the first
part and move to the second, which is the actual proposed solution]
On 08/11/2007, Guy Chapman aka JzG <guy.chapman(a)spamcop.net> wrote:
Here's a problem, though: there is a tendency to
assume bad faith on
the part of deleting admins,
(I don't think we're assuming bad faith so much as we're assuming
laziness... or that they assume good faith of the taggers too much!)
and not to address bad speedy tagging
by RC patrollers. I completely support any initiative to educate
those who patrol recent changes, to persuade them to make better use
of {{prod}} and {{afd}} rather than {{db}}.
Indeed. However, bad speedy tagging - through misunderstanding or
overenthusiasm - is inevitable, so our admins are always going to have
to stay on their toes and make judgements as to whether or not
deletion is appropriate. If we make the tagging better, maybe they'll
only have to discard 5% of them and not 50%. But there'll still be
discards or proddings or whatever, and there will be enough of them
they still need to treat it as likely.
Perhaps what is missing is a triage stage between RC
patrol
identifying a likely problematic article, and the deletion category.
Clearing CAT:CSD of a backlog of hundreds is fundamentally
incompatible with being truly diligent about anything.
The problem is that the process, as set up now, has a triage stage -
the deleting admin. It's categorised for deletion; the admin decides
to concur (delete), vacillate (pass to prod or AFD), or object
(detag). They're our decision-making process.
I'm not sure creating a specific triage "step" would achieve much.
Let's consider...
----
Say we create a decision-making stage between tagging and actually
listing for deletion in CSD. This stage would either: a) make
irrevocable decisions on delete/don't delete, so we could just nuke
anything they put in CSD; or b) do a first pass at weeding out the
obvious keeps, so we would still need to make decisions on the
material they put in CSD.
In the case of a), would we really need to have the deletion category?
If any material placed there is going to be deleted
near-automatically, we might as well just do the deletion rather than
relabelling it for imminent deletion; save twenty seconds. Even if
they don't press the button themselves, labelling for imminent
deletion is about as effective - and we thus have our triagers doing
the deleting step, whether by their own hand or by someone else's as
proxy, which is what we have now.
In the case of b), we'd end up with... well, with CSD as it is now,
split into two categories, one "filtered" and one not, but admins
would still have to be paying attention to the content of pages in the
filtered category.
Net result: either way, it is basically the same as CSD today.
----
As you can see, I'm not sold on adding another step to the
speedy-deletion process as being the solution - we're just going to
have to impress on people that they need to pay more attention to the
triaging, whatever way we go about things.
What might be a more productive approach is to look at disentangling
the two roles of CSD...
----
Currently, there are two types of material that go in CSD, at least in theory.
a) Patently obvious stuff, that no-one would argue we ought to keep
(and material we ought to keep that got mistagged) - speedy means
expedited, uncontentious, no process
b) Stuff that needs to be deleted *right now* - speedy means, well, *fast*.
(The latter is a subset of the former, but let's split them for the
moment - a) is the stuff that isn't actually harmful or damaging)
Because of the presence of b), and the fact the material is mixed
together with no distinction made between the classes, we have an
apparent sense of urgency to Get It Cleared, Get Them Deleted, Do It
Soon. But, in practice, the material in a) can linger - oh, we don't
want to keep it around for days on end, but it doesn't matter if it's
dealt with now or after lunch.
What makes this division interesting is that, in general, the "bad
speedies" - things that aren't really speediable but someone thought
they were - would fall into category a). if it seems harmful or
damaging, it almost certainly *ought* to be deleted, with a quick
glance at the history to make sure it isn't just a vandalised page.
How about we split CSD, to reflect these two sides?
i) Material for "urgent deletion"
ii) Material for "simple deletion"
Material in i) would be vandalism, attack pages, etc etc; ii) would be
all the gibberish and spam and vanity and dead redirects and other
housekeeping.
Critically, because of the nature of the "bad calls", discussed above,
we have a nice split whereby material in i) *can* easily be knocked
off quickly without needing to dig too far - because it is much more
likely to be legitimately deletable - and material in ii), which we
accept may get backlogged and may hang around for some time, is the
material that we are likely to need to spend more time looking at.
ii would tend to become a sort of "speedy PROD", I guess; it gets
tagged and might go in ten minutes, might be a day. Certainly this
sort of material conceptually resembles PROD a lot... the split
between CSD.ii and PROD is an open question. We might end up merging
the two approaches somehow, but that's another matter.
----
How does that sound? I suspect it would solve a lot of the problems
relating to real or presumed urgency, and would keep people from
feeling they had to move fast on disputable material. It would also
allow us to expedite dealing with the actually damaging stuff, which
is a plus in anyone's book.
--
- Andrew Gray
andrew.gray(a)dunelm.org.uk