Wikipedia:Wikipedia Signpost/2012-01-16/Technology report

ArticleFeedback 5 enters second trial phase


Having spent three weeks collecting data on which interface design prompts the most useful feedback, the ArticleFeedback version 5 trial entered its second phase this week. This phase will look at the impact of interface placement on the quality and quantity of feedback solicited from the small percentage of pages on which the extension will be active. As Oliver Keyes, a WMF Community Liaison helping with the tool : Even with Wikipedia readership reaching half a billion users per month, the feedback form its current position (at the end of the article) doesn’t see a whole lot of activity. In this test, we’ll be experimenting with a more prominent way to access to tool. When a user loads the page with the test version of the Article Feedback Tool, they will see an "Improve this article" link docked on the bottom right hand corner of the page.

The introduction of this link will undoubtedly increase the amount of feedback. We need to, however, understand how it affects the quality of the feedback. ... As with the last tests, it'll be on a very small subset of articles and probably won't be noticed by most people. ... We'll also be doing some preliminary analysis on whether such a prominent link cannibalizes editing behaviour.

It is possible for logged-in users to hide the display of both the current version 4, and the new version 5, of the ArticleFeedback extension via their user preferences; an RfC is also currently open for editors wishing to influence the future direction of the extension, particularly with regards to handling textual feedback such as general comments about the article. As already noted, it is planned that phase 3 will focus on the impact of the extension on editing levels. Phase 4 is set to focus on the impact of the tool on readers; phase 5 will look at its longer term impact on editor levels. Although versions of the extension are trialled on the English Wikipedia, other wikis also run versions of it.

In brief
Not all fixes may have gone live to WMF sites at the time of writing; some may not be scheduled to go live for many weeks.
 * Feedback dashboard improved: The FeedbackDashboard, currently used to display user comments made via the MoodBar extension, underwent an upgrade this week. According to a post, "MoodBar feedback posts are now listed within a user's contributions log; we have added a "Top responders" leaderboard to the Dashboard; [and] there is now an "Unanswered" filter for feedback dashboard". A number of more minor changes were also deployed.
 * Harassment policy drafted: Volunteer Development Coordinator Sumana Harihareswara described how she had begun drafting an official anti-harassment policy specifically targeted at "tech conferences and hackathons" but with potential application to all WMF events (wikitech-l mailing list). Operations Engineer and partial critic Peter Youngmeister noted that such a policy is useful "when you either a) want to kick someone [whose] actions are unacceptable out of your event or b) something bad happened and the organizer wants to be able to point to the policy and say 'that was against our policy.'" Most of the feedback received since the mailing list post has been positive, suggesting that the policy could come into force as early as this week at the San Francisco hackathon (20–22 January).
 * Wikimedia Tech receives interns: The WMF is to benefit from the expertise and time of a number of part-time developers, according to an announcement on the wikitech-l mailing list. According to the announcement, the "four bright college students already know web development and have had internships at Facebook, Google, and Microsoft... it's a bit like Google Summer of Code, but [with] a part-time team rather than one full-timer".
 * 1.18 upgrade "most difficult MW upgrade" for external sites: A post by DanB noting the multitude of issues he faces upgrading a large corporate wiki from MediaWiki 1.17 to MediaWiki 1.18 triggered a discussion about how problems such as these could be avoided in the future (wikitech-l mailing list: 1, [2]). As DanB noted, the purpose of his post was to be "constructive & helpful, not blameful: ... I'd just like to describe what kinds of things broke for a reasonably active wiki run by well-meaning people, and to document how we fixed them".
 * Improving Facebook embedding: There was a discussion on the wikitech-l mailing list to improve the way Wikimedia links "embed" into Facebook, for example, by suggesting that the initial thumbnail image should be the one used in an infobox. However, whilst separate issues concerning language variants could be fixed, the problem with images seemed to be the result of Facebook not following its own rules on thumbnail selection, Daniel Friesen explained, making it a difficult issue to fix.
 * IPv6 and Wikimedia: Two separate threads were started this week on the English Wikipedia technical village pump to look at issues surrounding the impending global transition to from version 4 to version 6 of the Internet protocol (usually referred to as IPv4 and IPv6 respectively). The highlighted possible problems with rangeblocks, while the  related to a number of issues, including problems with the current CheckUser setup and how IP address user talk pages function.
 * PostgreSQL support improving: Developer Antoine Musso announced the creation this week of a new unit test interface backed by a PostgreSQL database layer, in addition to the existing interface which uses a MySQL backend. Since most conventional testing is done on top of MySQL (the current DBMS of choice for Wikimedia wikis), the new project is expected to drastically improve the number of bugs and regression caught before code goes live.