Wikipedia:Wikipedia Signpost/2011-02-07/Technology report

Technology report

Bugs, Repairs, and Internal Operational News

MediaWiki 1.17 deployment imminent

For most of its history, code for the MediaWiki software which forms the basis of WMF wikis and other sites was developed, packaged into sequential releases every three or four months, and then deployed. However, more recently, major changes have been deployed immediately (i.e. out of cycle), easing the pressure on set-piece deployments of code. On Tuesday, February 8, starting at 07:00 UTC, the latest major release of the MediaWiki software, version 1.17, will be released, giving WMF wikis around six months' worth of not-already-deployed code updates (Wikimedia Techblog). A release to other sites will follow soon after. The release represents a massive code review effort in the last two months to check all updates to the software (approximately 1200 of them) before they are released.

If the deployment goes well, it is unlikely that Wikimedians will notice much difference: most of the major features have already been implemented on WMF wikis. (Some interim disruption is nonetheless expected; for example, database dumps are to be stopped temporarily.) Among the major unreleased developments is the new ResourceLoader, designed to speed up page loading times, though it could cause JavaScript errors on less well maintained (smaller) WMF wikis. Other updates are less obvious: a full list, which includes already released developments, is also available. A number of other developments dependent on 1.17 will be added shortly after its release.

The regularity of releases – and particularly the criteria used for determining which updates were deployed immediately and which had to sit in the queue – has been a contentious issue in recent months (see, for example, previous Signpost coverage from October 2010: 1, 2). There is now hope that after the release of 1.17, it may be possible to act upon volunteer developers' calls to have a more regular development cycle. For example, Brion Vibber, for a long time Chief Technology Officer (CTO) at Wikimedia, commented (wikitech-l mailing list):

February Engineering Update published

In addition to the major announcement concerning 1.17, the Foundation's Engineering Update for February (and covering the activities of January) was published last week on the Wikimedia Techblog, giving a brief overview of all Foundation-sponsored technical operations in the last month. It summarised the developments:

The update also noted that new job openings for the positions of Operations Engineer and Senior QA Engineer, previously announced, had been published; and that work had been done on evaluating LiquidThreads but that development of the Article Feedback tool had been put off until after the release of 1.17. Other noticeable improvements include the imminent launch of improved category collation code, allowing sub-categories, files, and pages to be paged separated. This "dark-launch" will be invisible at first, while the feature is stress-tested to check for errors before it is finished and made visible to the average user. According to the update, a survey related to the Wikimedia mobile site is also being drafted.

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.

  • After the news that all remaining IPv4 addresses had been allocated to regions – effectively signalling the possible end of the standard in some regions within the next eighteen months – there was a discussion about how Wikimedia sites could (better) support the new specification, IPv6 (wikitech-l mailing list).
  • The search engine Google briefly stopped including much of the German Wikipedia in their search listings after a "bug on their end" (Wikimedia bug #27155). There had been conjectures that it was related to the FlaggedRevs extension (which includes a "noindex,nofollow" tag in unreviewed page revisions, though by default these are only displayed as current revisions to logged-in users), which even made it into German media: a Spiegel online headline read "Bug lets Wikipedia vanish".
  • The Foundation held a "data summit" on February 3/4, "a working session about semantic data, analytics and research into data dumps" in California. Speakers included developers for Semantic MediaWiki and various projects that extract structured data from Wikipedia (including Freebase, DBPedia, Ultrapedia and Shortipedia), as well as in-house experts from Wikimedia such as Data Analyst Erik Zachte. The session on analytics included a one-hour discussion of privacy issues, with Foundation staff explaining potential community concerns to external researchers, e.g. about a possible use of a session cookie to track the reading behavior of Wikipedia visitors or analysis of the geographical provenance of logged-in edits. Videos from the event were streamed live (wikitech-l mailing list) and recordings are expected to become available online. A day later, the FOSDEM conference in Belgium featured a talk by Tomasz Finc and Arthur Richards titled "Free Culture, Free Data - How we use Data to Drive at Wikipedia".