[WikiEN-l] a valid criticism

Fri Oct 7 23:57:18 UTC 2005

On Fri, 7 Oct 2005, Magnus Manske wrote:

> Jimmy Wales wrote:
>
> >
> >But it's really not joyful to see bad writing in two very prominent
> >articles.
> >
> >
> Just as a reminder, there *is* a validation/rating/whatever-you-call-it
> system sitting in MediaWiki for month, waiting to be tested.
>
I'm attaching this to Magnus' email not so much as a response, but to
throw out an idea for allowing us to put this into testing, & hopefully
production.

As I understand, the issue why this feature has not been enabled on
Wikipedia is because of the load it would put on the servers: keeping
track of another mesh of tables with the validation values for every
revision of articles would bring the system to its knees.

What about only allowing users to rate *ONE* version of an article at
a time. If someone makes changes to an article, & you think it decreases
the quality, you can keep your rating on the older version; if it
improves the article, you move your rating to the lastest version. And
users can monitor all of this by using their watchlist function.

The one drawback I see with this (besides that it might not solve the
problem with the processors) is that contributors can only watch so many
articles, due to our own grey processor limits: while I've heard of folks
having thousands of articles on their watchlists, I suspect more people
are like me, with less than 200 articles watchlisted. So if the
practical limit is that users can only rate 400 articles (to pick a number)
at a time, & we have 400-600 active users rating articles, then there
will be between 160,000 - 240,000 ratings. While that may look like
enough to cover a good chunk of Wikipedia's 750,000+ articles, we need
a minimum of ratings on each article to really make this system work:
if we need a minimum of 5 ratings (again, to pick a number), we will
have only 32,000 - 48,000 articles with ratings.

But we'd have 5% - 7% of our articles with some kind of rating, which
is more than the approximately 0.15% of articles we've labelled FA, & it
would allow us to at least start on this problem.

Geoff