Hi all,
as promised on the meeting, I wrote a small JavaScript/Toolserver setup that can log clicks on Commons leading to external pages.
This requires JavaScript to be active, but that should be the case with the "normal" (read: not-geek) population. If JavaScript is turned off or not available, it degrades gracefully (it just won't count the click).
JavaScript : http://commons.wikimedia.org/wiki/User:Magnus_Manske/log_external_link_click... To activate : addOnloadHook ( log_external_link_clicks ) ;
The latter could go into the main Commons javascript page. No further setup is required.
This will "capture" all external link clicks within the page content, tell the toolserver about it, and then load the intended page.
A logfile will be generated on the toolserver, containing anonymized data like the Commons page and namespace, target URL, link text, timestamp etc. It also records content and user language, as well as the information if the user was logged in or not (but no user name or IP!).
Log files will be visible on the toolserver in /mnt/user-store/mm6_logs (1 file per month; can switch to 1 per day if volume is too large).
That should give us some nice statistics about how many people follow which link from where, which will be useful as an argument on GLAMs.
Note that with the same mechanism, I could log other events as well, e.g., clicks on an image on the image page (to see the full-size version). However, it would be overkill to use it for "image pages viewed" in general, IMHO.
Cheers, Magnus
I had the same thought when I first saw it.
On Wed, Nov 18, 2009 at 5:33 PM, Tim Starling tstarling@wikimedia.orgwrote:
Magnus Manske wrote:
Hi all,
as promised on the meeting, I wrote a small JavaScript/Toolserver setup that can log clicks on Commons leading to external pages.
I believe this would be a violation of Wikimedia's privacy policy.
-- Tim Starling
Commons-l mailing list Commons-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/commons-l
Reading the Privacy policy I dont see this would violate the policy as it specifically doesnt record private data(ip, username), in my reading of http://meta.wikimedia.org/wiki/Privacy_policy#Purpose_of_the_collection_of_p... policy specifically says " *The Foundation limits the collection of personally identifiable user data to purposes which serve the well-being of its projects...*." then includes some examples like public accountability of the projects, and site statistics(raw data is not made public). This the tool appears to be within those bounds of the privacy policy.
2009/11/19 Dakota sandahlb@gmail.com
I had the same thought when I first saw it.
On Wed, Nov 18, 2009 at 5:33 PM, Tim Starling tstarling@wikimedia.orgwrote:
Magnus Manske wrote:
Hi all,
as promised on the meeting, I wrote a small JavaScript/Toolserver setup that can log clicks on Commons leading to external pages.
I believe this would be a violation of Wikimedia's privacy policy.
-- Tim Starling
Commons-l mailing list Commons-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/commons-l
Commons-l mailing list Commons-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/commons-l
On Thu, Nov 19, 2009 at 9:23 AM, Gnangarra gnangarra@gmail.com wrote:
Reading the Privacy policy I dont see this would violate the policy as it specifically doesnt record private data(ip, username), in my reading of http://meta.wikimedia.org/wiki/Privacy_policy#Purpose_of_the_collection_of_p... the policy specifically says "The Foundation limits the collection of personally identifiable user data to purposes which serve the well-being of its projects...." then includes some examples like public accountability of the projects, and site statistics(raw data is not made public). This the tool appears to be within those bounds of the privacy policy.
Even if they would fall under the privacy policy, according to it the collection of private data can be done "including but not limited to the following" purposes: <snip> "To provide site statistics. The Foundation statistically samples raw log data from users' visits. These logs are used to produce the site statistics pages; the raw log data is not made public."
So in my opinion even if we _would_ save private information with the clicks, that still would not violate the privacy policy as long as the data is sufficiently anonymized before it is seen by human eyes.
2009/11/19 Tim Starling tstarling@wikimedia.org:
Magnus Manske wrote:
Hi all,
as promised on the meeting, I wrote a small JavaScript/Toolserver setup that can log clicks on Commons leading to external pages.
I believe this would be a violation of Wikimedia's privacy policy.
If it contains no explicitly private information (no username) and isn't linked to any other log entries (no "person 489272 clicked the following links...") - and from the way Magnus describes it this seems to be the case - then it seems to me like it wouldn't, because no data can be linked to a user in any reasonable or consistent manner. That said, IANAL. ;-)
(It is pushing into the area where we do need to be aware of privacy concerns, though!)
Hi everyone,
Magnus Manske schreef:
This will "capture" all external link clicks within the page content, tell the toolserver about it, and then load the intended page.
Something else: Is the toolserver able to handle the load this will produce? Maybe make it part of http://stats.wikimedia.org/ ?
Maarten
On Thu, Nov 19, 2009 at 1:08 PM, Maarten Dammers maarten@mdammers.nl wrote:
Hi everyone,
Magnus Manske schreef:
This will "capture" all external link clicks within the page content, tell the toolserver about it, and then load the intended page.
Something else: Is the toolserver able to handle the load this will produce? Maybe make it part of http://stats.wikimedia.org/ ?
I have no idea about the amount of traffic. I'd guess it's less than 1% of Commons page views, but I might be widely off.
One way to find out would be to turn it on for everyone :-)
But, being the script author, I'd rather have someone else put it into the default Commons JavaScript. IMHO it will work technically, I just want to avoid "developers do whatever they want on Commons"-issues. Volunteers?
Cheers, Magnus