document updated 16 years ago, on Jul 15, 2008
Some data I'd like to collect, on an ~hourly basis:
- What periods of time Conservapedia's HTTP server is totally unresponsive, as well as times when it's responding, but you have to reload 5 times to get a page
- [reason] it's been down quite a bit lately, I'm curious if there's any kind of trend
- When "night edit" mode is enabled
- When user creation is enabled.
- [reason] again, seems to be almost random
- [technical] [[Special:Userlogin]] will have a "Create an account" link when it's enabled
- [technical] also, you can get some data from [[Special:Log/newusers]] as well (probably best to combine this with the above)
- The number of edits per day, and the net-bytes-added, per namespace, per day.
- [reason] an attempt to get a more accurate gauge of how many positive contributors they have
- [technical] [[Special:Recentchanges]]
- [reason] reports will be similar to the various edit counters I did at Wikipedia
- BONUS: if I can get the net-bytes-added metric across different wikis (RationalWiki, Uncyclopedia, and Wikipedia, for instance), that'd be neato-keen. Fortunately, api.php appears to work on all of them.
In addition, I've had several requests to resuscitate my "tool1" edit counter. I should probably rewrite the backend to get the data from api.php instead, since all wikis I use seem to have that now.
- it should cache heavily
- it can use list=users usprop=editcount to quickly check whether the cache needs to be discarded or not
- it can pull the original data from list=usercontribs
- detailed requirements:
- it should allow ANY minor mediawiki to add itself to the list of available wikis, just by asking for "the link to the history page for ANY article"... this lets it know where index.php is, which will let it know where api.php is, if it exists
- if the HTTP referer is in its list of wikis it knows about, it'll make that one the default in the drop-down. (however, for users who block referer, it'll have a CGI-parameter way to specify it as well)
- when it starts to load new data for a user, the CGI process should detach (so that the HTTP request is finished), and then the page should do long polling to wait for the response to come back
- also, if a new request is made for data to be processed, when the data-download is already in progress, then it should appear to "resume"... the original "time processing" should start at the current time-elapsed, and it should continue polling as if nothing ever stopped