On 8/21/06, Bryan Derksen bryan.derksen@shaw.ca wrote:
My support for the whole endeavour, for one.
[snip]
I hope you understand that getting the permission of each and every user before making a change is a solution which doesn't scale.
We've lost a lot of time as well. It's been almost nine months now, during some of the most widespread publicity Wikipedia's ever had. If when analysis is finally published it turns out that this experiment had a major negative impact on new user participation it would have been nice to know that more quickly to limit the damage.
I don't believe we could have gotten useful data about the possible positive impacts of this change any sooner than now. The most useful metrics take time (i.e. life expectancy of an article), and there is a lot of noise.. so to make confident statements we need lots of data.
For your reference, there have been people watching this for clearly negative impacts all along. Had there been substantial evidence that it was doing a significant amount of harm, it would have been aborted.
Here is some data, http://myrandomnode.dyndns.org/new_articles.png The graph shows the rate of article creation, excluding articles which were eventually deleted, based on the timestamp of the oldest visible revision over the life of the project. X is time in julian day numbers, the ticks are about 200 days apart, and the graph is about a month old.
This doesn't tell us that the policy *helped* but it does indicate that it isn't causing serious harm. A more comprehensive statistical study of the impact is forthcoming.
From a purely data standpoint probably not a whole lot was lost thanks to the thoroughness of the database's records, though having to rely on the deletion database for long-term information retention does make me a bit nervous. But that's not the only consideration here.
Our retention of deleted metadata is just as robust as anything else...