On 12/1/05, Daniel Mayer <maveric149(a)yahoo.com>
wrote:
What we need are teams of people working together
in a way that minimizes duplicated effort. As
is, there is no way to know if a diff has been looked at once before, never before, or
200 times.
There is also no way to know what was being looked at (simple vandalism, subtle
vandalism,
accuracy, etc). Knowing at least the first part (how many times a diff has been looked
at) will
help a great deal. Sorting RC based on that would be even better.
Better still would be to sort RC and similar lists based on trust networks. For example,
I trust
you and many other people. So edits either made by you and diffs checked by you should be
OK in my
eyes. It would be nice if my RC and similar lists de-emphasized your edits and edits
checked by
you. This could go one step farther: I trust your ability to judge whether or not other
people are
trustworthy. So it would be nice if I could trust by proxy all the users you trust. This
creates a
trust network.
This could also be done P2P via an offline editor and thus minimize server load. The
offline
editor would do all the sorting after downloading raw data from Wikipedia.
I think it is obvious that we need bigger and better guns to fight vandals.
This "trust network" sounds not entirely different to the whitelist on
CDVF. It shouldn't be too difficult to create a shared whitelist that
members of a specified group can all use and add to. Multiple lists
could be created and you could subscribe to as many as you like and
that the members accept you.
This sounds eminently possible to me, if a fair bit of work. Do
people with some technical knowledge have an idea if it is possible?
--
Sam
It's posible but it hits a brick wall pretty fast. A lot of the
vandalism on my watchlist is from anons.
Optimiseing is good but it hits problems:
You can only go so far.
If you sart ignoreing too many edits because you think they are safe
vandles will start to exploit that.
--
geni