Hi,
Looking at the summary reports per language, I've noticed a linear, significant increase in pageviews for many European languages (ro, bg, hu, fr) Wikipedias in the last 3 months. This is not happening for Asian languages or Russian and is not obvious from the report card.
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Thanks, Strainu
Hi Strainu,
We noticed and are investigating. It surely looks almost too good to be true. Any suggestions for an explanation are welcome.
Cheers, Erik
-----Original Message----- From: wikimedia-l-bounces@lists.wikimedia.org [mailto:wikimedia-l-bounces@lists.wikimedia.org] On Behalf Of Strainu Sent: Friday, November 22, 2013 22:41 To: Wikimedia Mailing List Subject: [Wikimedia-l] Increase in page views for the last 3 months
Hi,
Looking at the summary reports per language, I've noticed a linear, significant increase in pageviews for many European languages (ro, bg, hu, fr) Wikipedias in the last 3 months. This is not happening for Asian languages or Russian and is not obvious from the report card.
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Thanks, Strainu
[1] http://stats.wikimedia.org/EN/SummaryRO.htm
_______________________________________________ Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
I would guess either national press, or a change in how the popular search engines work?
Are we running & recording certain defaulted random queries from time to time to compare?
Lodewijk
2013/11/22 Erik Zachte ezachte@wikimedia.org
Hi Strainu,
We noticed and are investigating. It surely looks almost too good to be true. Any suggestions for an explanation are welcome.
Cheers, Erik
-----Original Message----- From: wikimedia-l-bounces@lists.wikimedia.org [mailto: wikimedia-l-bounces@lists.wikimedia.org] On Behalf Of Strainu Sent: Friday, November 22, 2013 22:41 To: Wikimedia Mailing List Subject: [Wikimedia-l] Increase in page views for the last 3 months
Hi,
Looking at the summary reports per language, I've noticed a linear, significant increase in pageviews for many European languages (ro, bg, hu, fr) Wikipedias in the last 3 months. This is not happening for Asian languages or Russian and is not obvious from the report card.
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Thanks, Strainu
[1] http://stats.wikimedia.org/EN/SummaryRO.htm
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Hi,
I closely monitor Indic WP stats and witnessed a similar near 100% spike during April to June 2013. http://stats.wikimedia.org/EN/SummaryBN.htm
Tried very much to find out what triggered this but didn't come across anything that significant to have caused this. Wondering if there is any way to get a more nuanced understanding of such events? It would be helpful.
Vishnu
On Saturday, 23 November 2013, Lodewijk lodewijk@effeietsanders.org wrote:
I would guess either national press, or a change in how the popular search engines work?
Are we running & recording certain defaulted random queries from time to time to compare?
Lodewijk
2013/11/22 Erik Zachte ezachte@wikimedia.org
Hi Strainu,
We noticed and are investigating. It surely looks almost too good to be true. Any suggestions for an explanation are welcome.
Cheers, Erik
-----Original Message----- From: wikimedia-l-bounces@lists.wikimedia.org [mailto: wikimedia-l-bounces@lists.wikimedia.org] On Behalf Of Strainu Sent: Friday, November 22, 2013 22:41 To: Wikimedia Mailing List Subject: [Wikimedia-l] Increase in page views for the last 3 months
Hi,
Looking at the summary reports per language, I've noticed a linear, significant increase in pageviews for many European languages (ro, bg,
hu,
fr) Wikipedias in the last 3 months. This is not happening for Asian languages or Russian and is not obvious from the report card.
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Thanks, Strainu
[1] http://stats.wikimedia.org/EN/SummaryRO.htm
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Erik Zachte, 22/11/2013 23:21:
We noticed and are investigating. It surely looks almost too good to be true. Any suggestions for an explanation are welcome.
To address wild speculations on https://en.wikipedia.org/wiki/Knowledge_Graph , it would be quite useless to have even just a simple thing as an archive of all updates to http://stats.wikimedia.org/wikimedia/squids/SquidReportGoogle.htm . I hope the server has space for 130 KB more each month. :)
Nemo
I meant "useful", of course; sorry for double post.
Federico Leva (Nemo), 23/11/2013 09:26:
Erik Zachte, 22/11/2013 23:21:
We noticed and are investigating. It surely looks almost too good to be true. Any suggestions for an explanation are welcome.
To address wild speculations on https://en.wikipedia.org/wiki/Knowledge_Graph , it would be quite useless to have even just a simple thing as an archive of all updates to http://stats.wikimedia.org/wikimedia/squids/SquidReportGoogle.htm . I hope the server has space for 130 KB more each month. :)
Nemo
On 11/22/2013 01:41 PM, Strainu wrote:
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Pretty similar to Spanish and Catalan:
http://stats.wikimedia.org/EN/SummaryES.htm http://stats.wikimedia.org/EN/SummaryCA.htm
Some Catalan editors {{vague}} were wondering how much the nice pannel featuring Wikipedia text and image in Google searches had to do with this. But yes, it's almost too nice to be true.
It would be interesting to see whether the increase in page views pulls a vawe of increased edits and editors numbers.
Similar thing to Gujarati Wikipedia http://stats.wikimedia.org/EN/SummaryGU.htm
Peak at August 2013 Down at April 2013
It may be possible that people who are using local language google search increased.
On Sat, Nov 23, 2013 at 12:25 PM, Quim Gil qgil@wikimedia.org wrote:
On 11/22/2013 01:41 PM, Strainu wrote:
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Pretty similar to Spanish and Catalan:
http://stats.wikimedia.org/EN/SummaryES.htm http://stats.wikimedia.org/EN/SummaryCA.htm
Some Catalan editors {{vague}} were wondering how much the nice pannel featuring Wikipedia text and image in Google searches had to do with this. But yes, it's almost too nice to be true.
It would be interesting to see whether the increase in page views pulls a vawe of increased edits and editors numbers.
-- Quim Gil Technical Contributor Coordinator @ Wikimedia Foundation http://www.mediawiki.org/wiki/User:Qgil
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
I have assumed it is an effect of Google starting to show an extract from Wikipedia on the page where they show hit results.
I raised the issue Sept 29 in a thread here called New Google interface to Wikipedia
Anders
Harsh Kothari skrev 2013-11-23 08:00:
Similar thing to Gujarati Wikipedia http://stats.wikimedia.org/EN/SummaryGU.htm
Peak at August 2013 Down at April 2013
It may be possible that people who are using local language google search increased.
On Sat, Nov 23, 2013 at 12:25 PM, Quim Gil qgil@wikimedia.org wrote:
On 11/22/2013 01:41 PM, Strainu wrote:
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1]. [1] http://stats.wikimedia.org/EN/SummaryRO.htm
Pretty similar to Spanish and Catalan:
http://stats.wikimedia.org/EN/SummaryES.htm http://stats.wikimedia.org/EN/SummaryCA.htm
Some Catalan editors {{vague}} were wondering how much the nice pannel featuring Wikipedia text and image in Google searches had to do with this. But yes, it's almost too nice to be true.
It would be interesting to see whether the increase in page views pulls a vawe of increased edits and editors numbers.
-- Quim Gil Technical Contributor Coordinator @ Wikimedia Foundation http://www.mediawiki.org/wiki/User:Qgil
Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
Anders Wennersten, 23/11/2013 08:49:
I have assumed it is an effect of Google starting to show an extract from Wikipedia on the page where they show hit results.
Assumptions are dangerous. The feature was apprently enabled on 2012-12-04 for Spanish, French, German, Portuguese, Japanese, Russian and Italian; all those languages show a decrease in page views on that month (except Russian whose trend is constant though)... http://stats.wikimedia.org/EN/TablesPageViewsMonthlyCombined.htm
Nemo
Thanks all for thinking along.
We have found the cause of the unbelievable growth in page views, and it turns out to be an bug indeed.
Around August 2013 a site change caused internal housekeeping messages to be counted as page views by our webstatscollector software. As the patch was rolled out progressively, every month more bogus page views were added, up to several billion per month in November.
All page review reports now have a very clear warning about this issue. http://stats.wikimedia.org/EN/TablesPageViewsMonthlyCombined.htm
Thanks to Christian Aistleitner for pinpointing the specific new url's that caused overcount, of up to 100 million per day in November:
Special:CentralAutoLogin/createSession Special:CentralAutoLogin/start ../autonym.ttf
Knowing this we can patch the hourly projectcount files. Original files: http://dumps.wikimedia.org/other/pagecounts-raw/2013/2013-11/ Patched files: http://dumps.wikimedia.org/other/pagecounts-ez/projectcounts/
I'll reply on this thread when the patch has been applied.
Erik Zachte
On Wed, Nov 27, 2013 at 1:00 PM, Erik Zachte ezachte@wikimedia.org wrote:
Special:CentralAutoLogin/createSession Special:CentralAutoLogin/start
You should also remove anything else beginning with Special:CentralAutoLogin/.
Maybe Special:MWOAuth/ and Special:OAuth/ too.
Brad Jorsch (Anomie), 27/11/2013 20:38:
On Wed, Nov 27, 2013 at 1:00 PM, Erik Zachte ezachte@wikimedia.org wrote:
Special:CentralAutoLogin/createSession Special:CentralAutoLogin/start
You should also remove anything else beginning with Special:CentralAutoLogin/.
Maybe Special:MWOAuth/ and Special:OAuth/ too.
Or they could be used via /w/index.php* URLs so that they're not counted?
Nemo
Hi Brad,
On Wed, Nov 27, 2013 at 02:38:04PM -0500, Brad Jorsch (Anomie) wrote:
On Wed, Nov 27, 2013 at 1:00 PM, Erik Zachte <ezachte at wikimedia.org> wrote:
Special:CentralAutoLogin/createSession Special:CentralAutoLogin/start
You should also remove anything else beginning with Special:CentralAutoLogin/.
Maybe Special:MWOAuth/ and Special:OAuth/ too.
Thanks for pointing that out.
I have not fully caught up on reading up on the internals of our OAuth implementation, but when looking at the sampled 1:1000 logs, I could only find two requests to
Special:OAuth/initiate
. No other requests to Special:MWOAuth/ , Special:OAuth/ . Should we already see traffic to those endpoints?
Best regards, Christian
On Sat, Nov 30, 2013 at 6:20 PM, Christian Aistleitner christian@quelltextlich.at wrote:
. No other requests to Special:MWOAuth/ , Special:OAuth/ . Should we already see traffic to those endpoints?
Not a whole lot, yet, and it may never really get to be *that* many. On the other hand, maybe it will. It certainly shouldn't ever get anywhere near the level of Special:CentralAutoLogin.
The /initiate you saw should have been followed by a call to /authorize and then likely a call to /token, but of course those could have missed the sampling. We've also got /verified, /identify, and /grants in there.
Of all these, /authorize, /verified, and /grants are the only ones that could remotely be considered a real pageview. You can see /authorize by going to https://tools.wmflabs.org/oauth-hello-world/enduser.php and clicking the "Make an edit" button, /verified at https://www.mediawiki.org/w/index.php?title=Special:OAuth/verified&oauth..., and /grants at https://www.mediawiki.org/wiki/Special:OAuth/grants.
Hi Brad,
On Mon, Dec 02, 2013 at 10:18:08AM -0500, Brad Jorsch (Anomie) wrote:
On Sat, Nov 30, 2013 at 6:20 PM, Christian Aistleitner christian@quelltextlich.at wrote:
. No other requests to Special:MWOAuth/ , Special:OAuth/ . Should we already see traffic to those endpoints?
Not a whole lot, yet, and it may never really get to be *that* many. On the other hand, maybe it will. [...]
Ok. Thanks for the clarification.
The /initiate you saw should have been followed by a call to /authorize and then likely a call to /token, but of course those could have missed the sampling.
Yes, probably.
Thanks for explaining the course of actions for the endpoints, and the examples.
Best regards, Christian
It's not clear if this is a bug or true organic growth, but it seems to be occurring across multiple Wikipedias (see the rest of the thread).
Matt Flaschen
-------- Original Message -------- Subject: [Wikimedia-l] Increase in page views for the last 3 months Date: Fri, 22 Nov 2013 23:41:09 +0200 From: Strainu strainu10@gmail.com Reply-To: Wikimedia Mailing List wikimedia-l@lists.wikimedia.org To: Wikimedia Mailing List wikimedia-l@lists.wikimedia.org
Hi,
Looking at the summary reports per language, I've noticed a linear, significant increase in pageviews for many European languages (ro, bg, hu, fr) Wikipedias in the last 3 months. This is not happening for Asian languages or Russian and is not obvious from the report card.
Has anything changed in the reporting or the visit patterns for these Wikipedias? It looks pretty weird to have a 100% increase for Romanian in just 3 months [1].
Thanks, Strainu
[1] http://stats.wikimedia.org/EN/SummaryRO.htm
_______________________________________________ Wikimedia-l mailing list Wikimedia-l@lists.wikimedia.org Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, mailto:wikimedia-l-request@lists.wikimedia.org?subject=unsubscribe
wikimedia-l@lists.wikimedia.org