All dumps ran from the data center in Ashburn this month; only the miscellaneous and experimental services remain to be moved. GSOC student Petr Onderka completed the first incremental dump-producing code, along with a draft specification for the new format. Test it out and let us know what you think!
Due to Wikimania and staff vacations, this month had a relatively low number of infrastructure changes, but we had a relatively high influx of users and tools. We ran three workshops during Wikimania and helped Toolserver users migrate their tools to Labs. We did have a few infrastructure changes, though: A change for the service group interface was merged but not yet deployed. It removes the service group interface from the project interface, reducing clutter. An API was pushed in for project and service group information, to make the information available from Wikitech, rather than LDAP. Other infrastructure changes were bugfixes, which can be found through bugzilla.
OTRS got a long overdue update to version 3.2.9 with the generous volunteer support of Martin and Marcel of Znuny GmbH. As part of the upgrade, the service was migrated from pmtpa to eqiad, and spam filtering was overhauled.
In August, the VisualEditor team continued work, and presented and ran workshops at Wikimania in Hong Kong to discuss how to best improve the system. The deployed version of the code was updated three times (1.22-wmf13, 1.22-wmf14 and 1.22-wmf15), with several mid-deployment releases as the code was developed to patch urgent issues. The focus in this work was in improving the stability and performance of the system, fixing a number of bugs uncovered by the community, and making some usability improvements.
In August, the Parsoid team continued to polish compatibility with existing wikitext. User feedback after the July VisualEditor release was instrumental in the identification of issues and the development of support for important use cases of creative templating.
The increased team size also allowed us to perform some long-standing code cleanup, make Parsoid compatible with Node 0.10, and improve testing. The round-trip testing infrastructure received a much-needed overhaul. The storage back-end switched from SQLite to MySQL, which improved throughput a lot and is allowing us to test new code far more quickly than before. Performance statistics are now recorded, which will let us identify performance bottlenecks as well as catch performance regressions.
In August, we released Notifications on the French, Hungarian, Polish, Portuguese and Swedish Wikipedias, after extensive testing on the English Wikipedia, as well as mediawiki.org and Meta-Wiki. This engagement tool was well received by our new communities, especially social features such as Mentions and Thanks, which enable users to communicate more effectively than before. Benny Situ led the engineering work for this deployment and fixed a number of bugs, with the help of Erik Benhardson and Matthias Mullie. Fabrice Florin managed community relations for these new releases, updating this release plan and reaching out to more projects, to prepare for worldwide deployments on all wiki projects in coming months. To that end, we teamed up with Philippe Beaudette, Maggie Dennis, Patrick Earley, Jan Eissfeld, Anna Koval, Keegan Peterzell, and Sherry Snyder to coordinate these releases with the communities they serve. Dario Taraborelli created new metrics dashboards for French, Hungarian, Polish, Portuguese and Swedish Wikipedias. Lastly, we presented our work on Notifications in two talks at Wikimania 2013, with both a general overview and a technical presentation (see slides). We are very grateful to all our community champions for each language and look forward to more collaborations in the future. Our next major deployment to non-English Wikipedias will take place on Sep. 17, to be followed by weekly releases throughout the fall, as outlined in our release plan. To learn more, visit the project portal, read the help page and join the discussion on the talk page.
In August, we continued development of the Flow prototype by implementing revisioning, moderation, and display code, on top of the storage and block abstractions. We have deployed this prototype to an internal labs instance to encourage the full team's involvement in development. Additionally, we participated in an agile workshop run by Arthur and Tomasz from the mobile team. This workshop facilitated planning the Flow MVP and setting goals for the team's first development sprint, along with providing information about agile guidelines and practices that have worked well for the mobile team.
In August, we made a few feature tweaks and bug fixes for the Article Feedback Tool (AFT5) on the English and French Wikipedias. Matthias Mullie released a few patches to improve the opt-in/opt-out tool, and tested the new feedback notifications to let users know when feedback is marked as useful for a page they watch (or for a comment they posted). We also presented our work on AFT5 at Wikimania 2013, with designer Pau Giner and our French and German champions Benoît Evellin and Denis Barthel, in this session (see slides). The team plans to make the AFT5 tool available to other wiki projects interested in testing this tool later this year, as outlined in the release plan.
This month, the Mobile Apps team pushed out additional releases of the Commons photo uploader app for iOS and Android. The iOS version includes a major UI revamp by Monte, while the Android version has received multiple incremental updates by Yuvi and Brion. Yuvi has been working on modernizing support for campaigns in UploadWizard, which will make it easier to coordinate uploads for events like Wiki Loves Monuments. Viewer, contributor, and admin user interfaces for campaigns will come to the web, with campaign-tied uploading in the web and mobile app. The team also started making plans for the next generation of the Wikipedia reader app, which will be more closely integrated with the mobile web site to ensure that new features are always available through a web view, even where we don't create specific native support. More details will be put together in the next couple months.
This month, the team completed version 1 of Wikipedia Zero automation tests, continued programming the re-architecture of Wikipedia Zero, implemented search engine non-indexing, and analyzed HTTPS requirements in support of a push for greater usage of HTTPS across Wikimedia projects. The Wikipedia Zero engineering team thanks Amit Kapoor from the Wikipedia Zero partnerships team, who wrapped up work with Wikimedia Foundation this month, for his hard work getting the program off the ground. And the team is also pleased to welcome Carolynne Schloeder, who joins the Wikipedia Zero program as Director of Mobile, Programs.
This month we continued to improve the mobile editing feature, monitoring and triaging bugs and expanding the feature show at the section level of articles. We also released the first iteration of mobile notifications to projects where Echo is enabled (English, French, Polish, Portuguese, Hungarian, and Swedish Wikipedia, as well as and Meta). In beta, we built a new notifications treatment to be released in later months and continued working on mobile talk pages.
The team continued mentoring four Google Summer of Code (GSoC) students. Praveen Singh, mentored by Santhosh Thottingal, released a Chrome extension for Wikimedia Input Tools and contributed to the Indic Font Specification, a collaborative open source project. Team members also continued to work with Red Hat on various language initiatives.
In August, we continued to expand our multimedia team: Bryan Davis joined us as senior platform engineer, working with product manager Fabrice Florin, front-end engineer Mark Holmquist and engineering director Rob Lanphier. We discussed multimedia plans and new feature ideas with community members in two separate events: a multimedia roundtable at Wikimania 2013 and an IRC chat, and updated our multimedia plan for the coming year based on their feedback (see slides). Summer contractor Brian Wolff completed the development of new gallery tags to support more appealing layouts for thumbnails, while Jan Gerber made improvements to the Score and TimeMediaHandler extensions. Mark Holmquist started development on the Media Viewer, based on designs by May Tee-Galloway and Jared Zimmerman; this new tool will display images in larger size when clicking on article thumbnails, as well as display file information and a full-screen viewing option, right on the same page. We aim to test a first version of this tool as part of a beta experiment on a few pilot sites in coming weeks. To discuss these features and keep up with our work, we invite you to join this new multimedia mailing list. Last but not least, we are also recruiting for two more positions for our team: a multimedia systems engineer and a senior software engineer. Please spread the word about this unique opportunity to create a richer multimedia experience for Wikipedia and MediaWiki sites!
In August we deployed CirrusSearch to test2.wikipedia.org and mediawiki.org and we're testing there. We're actively looking for other volunteers to test out CirrusSearch. Right now, CirrusSearch is not the primary search for mediawiki.org; you have to use a URL parameter to test it. We're hoping to make it the primary in September.
This month QA began collaborating closely with Release Engineering to coordinate improvement of reporting, monitoring, and testing software releases. Our goal is to make our frequent software releases even more reliable than they already are, and to use the tools and systems in place today such as the beta labs cluster to make those reliable releases even more frequent.
This month saw a significant change to the structure and organization of browser tests, with tests and builds for CirrusSearch, UniversalLanguageSelector, and VisualEditor following the example of MobileFrontend and now residing in the git repositories for those extensions, rather than in the /qa/browsertests repository. This creates opportunities for more frequent and more accurate Jenkins builds of the tests, while also reducing the overhead required for analyzing test failures.
Guillaume Paumier continued to focus on the VisualEditor deployment effort, working on communications, documentation and liaising with the French Wikipedia. Work on technical communications mostly focused on perennial activities like Tech news and ongoing communications support to the engineering staff.
We continue to pursue the initiatives listed in our planning document. We've had one analyst accept a job offer (welcome Aaron!) and are in discussions with a software engineer. We continue to have a solid pipeline and are spending a lot of time interviewing. Wikimetrics is on target for an early September release and we've made good progress against our hadoop infrastructure goals. In co-operation with Ops, we've completed our reinstall of the Hadoop cluster and run several days of reliability testing over the labor day weekend. We are currently investigating replacing the Oracle JDK with the Open JDK to be in line with our goals of using open source whenever possible. Our project to replace udp2log with Kafka is making steadily progress. Varnishkafka, which will replace varnishncsa, has been debianized and the first performance tests of compressing the message sets are very encouraging. We created a test environment in Labs to test Kafka failover modes and we have been prototyping with Camus to consume the data from a broker and write it to HDFS. We are right now thinking about how to set up Kafka in a multi data-center environment. The Zookeepers have been reinstalled through Puppet as well.
In close collaboration with Dario, Jaime and Jessie, we have worked on new features for Wikimetrics. In particular, we are adding new metrics such as survival, pages created, aggregation of metrics, metadata in the CSV output, a support page and we have now more than 90% test coverage of the codebase. In preparation for the reinstallation of the Hadoop cluster, we moved all Wikipedia Zero jobs off the cluster. We took this opportunity to add additional monitoring to the creation of Wikipedia Zero dashboards. We have worked with Wikipedia Zero to identify a problem with Geolocation of requests that has created large jumps in total traffic. We spent quite some time creating a more robust process for updating and monitoring gp.wmflabs.org. This dashboard is used by various internal stakeholders and receives its information from different datastreams using different scripts. We have been working on running these scripts under the general purpose stats user, adding additional monitoring to prevent stale data and puppetized some of the jobs.
In August, we attended WikiSym and Wikimania. Dario Taraborelli gave a keynote address on actionable Wikipedia research at WikiSym, where several other Wikipedia research papers were presented. At Wikimania, we hosted two sessions focused on Wikimedia data and analytics tools. We also worked with Platform engineering this month on analyzing and visualizing HTTPS failure rates by country, in preparation for the switch to HTTPS as a default. We released new dashboards for the launch of notifications on 5 other Wikipedias and continued to provide ad-hoc support to teams in Editor Engagement. Last, we continued screening and interviewing candidates for an open research analyst position.
Release of the new Mediawiki offliner was a little bit delayed; we are still fixing stability bugs. This solution has already proven its efficiency, as we have released 20 new ZIM files this month: a new throughput record. The ZIM incremental update GSoC project progresses too, as the student works on the integration of zimdiff/zimpatch in the Kiwix ecosystem. Kiwix developers have had a 6 days hackathon in Hong-Kong to prepare the next Kiwix release, after some final work on compilation.
In August, the Wikidata team was present at 3 events: COSCUP, Wikimania and a meetup about Wikidata and Incubator. A lot of work has been put into improving the API and its documentation. The team also worked on the ability to reorder the qualifiers and sources in a statement, improved the speed of Wikidata slightly, and made progress on the ability to query for statements with a specific property and value, as well as merging items. An improved proposal for the support of Wiktionary has been published. They also started the paper cuts initiative to find and fix small bugs that have a large impact on how enjoyable it is to use Wikidata. Denny and Adam gave a short overview of the state of Wikidata and answered questions during an office hour on IRC. The biggest news for August though was the activation of data access (Wikidata phase 2) on Wikivoyage.
The engineering management team continues to update the Deployments page weekly, providing up-to-date information on the upcoming deployments to Wikimedia sites, as well as the engineering roadmap, listing ongoing and future Wikimedia engineering efforts.