On Sat, 2009-12-12 at 18:27 -0700, bawolff wrote:
Note, I'm 99% sure that pulling the rating data off NewsTrust using clientside js would be in violation of the wmf privacy policy (99% as i have not actually read the privacy policy...). It also represents somewhat of a security risk (the usual method of doing such things is to execute foreign javascript that inserts rating data and/or puts data in a global variable. which is generally something we try to avoid, as executing arbitrary code on the client that is not under our direct control = bad thing in my mind). With that being said, that would probably be ok as a gadget, but not ok as enabled for everyone.
However, if we were to do this, i think a better approach would be to get someone with a toolserver account to create a tool that acts as intermediatery. That way, only the toolserver is in the position to potentially collect user data.
It'd go like this: Client loads page, page asks toolserver what rating of source is, toolserver asks newstrust (possibly caching results).
I assume that'd take care of privacy issues
I do worry how much work that would impose on NewsTrust. There would need to be some sort of API on their end to serve requests up with needed data.
It would also make having a "mission critical" Toolserver box essential. I know how flaky the toolserve has historically been. We can't rely on what's there for stuff appearing in published main namespace content.
In any case, we'd need to be sending the following information to NewsTrust from the Toolserver:
article URL source name if an initial request, or periodic polling
It'd need to return
any rating they have for the article an indicator there rating is for the article the number of reviews for the article any rating they have for the source the number of reviews for the source
If a periodic polling, NewsTrust could return some sort of "no change" indicator.
This would be flexible enough that Wikinews could collect the information on our own articles and stay within the privacy policy and the WMF techies paranoia about cross-site scripting attacks.
For us using the NewsTrust review widget in the {{social bookmarks}} template Bawolff, is there any way we can pick out best-guess categories when someone submits a story? As I said, the "basic" submission from where we've got the code fails until you complete those details.