Wikimedia Engineering/Report/2011/October

Major news in October include:

The New Orleans hackathon, which focused on Wikimedia's infrastructure;
Native HTTPS support on all Wikimedia sites;
Progress on the Visual editor project, with the first prototype expected in the coming months;
The deployment of the Translate extension to meta-wiki;
The deployment of MediaWiki 1.18 to all Wikimedia sites;
The completion of the first revision of the MediaWiki architecture document;
The ramp-up by the fundraising engineering team, to prepare for the upcoming annual fundraiser.

Hover your mouse over the green question marks ([?]) to see the description of a particular project.

Events

Recent events

New Orleans hackathon (14–16 October, New Orleans, USA) — About 30 MediaWiki developers and friends attended this coding event (organized by Ryan Lane, Sumana Harihareswara and Dana Isokawa) to work on Wikimedia's technical infrastructure. Work focused on the virtualization cluster / Wikimedia Labs, media storage (using Swift), Puppet, and continuous integration. The event was also used to discuss and prepare for the upcoming migration to git as our primary version control system.

October 2011 Coding Challenge (20 October - 7 November, online) — Greg DeKoenigsberg continued to prepare for this contest, which officially started on October 20th, and will be running for a few more days (see the rules). The contest offers three challenges: uploading media via a smartphone, surfacing changes in real time, and showing slideshows. The back-end, which provides a sign-up and submission process, was developed by Jeroen De Dauw and released as a MediaWiki extension. You can still participate in the challenge.

Upcoming events

India hackathon (18–20 November, Mumbai, India) — Alolita Sharma, Siebrand Mazeland, Sumana Harihareswara and the local India team continued to prepare for this event, which will focus on language, mobile and offline support for MediaWiki content. You can register to request a free invitation; approximately 100-125 attendees are expected.

Brighton hackathon (19–20 November, Brighton, England) — Free registration opened for this general MediaWiki hackathon planned by Lewis Cawte. You can register online. WMF engineers Antoine Musso, Roan Kattouw, and Sam Reed plan to attend.

San Francisco event (21-22 January 2012) — Erik Moeller and Sumana Harihareswara worked on planning for an outreach-focused developers' weekend. Staff and volunteer developers will help interested technologists learn to make awesome things with Wikimedia, by improving the Wikimedia platform, creating new functionality for Wikimedia users, or integrating Wikimedia content into their own products and services.

Personnel

Job openings

Are you looking to work for Wikimedia? We have a lot of hiring coming up, and we really love talking to active community members about these roles.

One new position opened in October: Software Security Engineer, and an RfP was published: RFP/Mobile UI/UX Redesign.

The following positions are still open:

The following requests for proposals are still open:

Short news

Phil Chang joined the Mobile team as Product Manager (announcement).
Leslie Carr joined the Operations Team as Operations and Network Engineer (announcement).
Antoine Musso joined the MediaWiki Core team as a contractor working on Continuous integration (announcement]).
Amir Aharoni joined the Internationalization and localization team as Software developer (announcement).
Two contractors joined the Strategic Product team: Fabrice Florin, leading the Article feedback 5.0 feature, and Oliver Keyes, acting as Community liaison (announcement).
Gabriel Wicke joined the Features engineering team as Software developer, focusing on the Visual editor (announcement).
Yoni Shostak (project manager), Greg Chiasson (developer), Reha Sterbin (developer) and Sean Heavey (UI designer/front-end developer), from OmniTI, will be working on the Article Feedback 5.0 feature.

Operations

Site infrastructure

Data Centers [?] — Chris Johnson, Rob Halsell and Peter Youngmeister have racked and brought online new servers in Tampa to enhance the capacity of the Bits, MediaWiki and API application servers. Leslie Carr worked on partitioning with PartMan and documented the process. An external firm performed a security audit on our infrastructure on our behalf. The team also addressed several high priority break fixes, like the external data storage failures, and the analytics packet loss due to heavy load with udp2log processing. Peter Youngmeister upgraded all Apache servers to Ubuntu 10.04; Squid servers have also been upgraded, and the Linux kernels have been updated with the latest security patches. Ben Hartshorne continued to set up servers for external storage. Mark Bergsma and Asher Feldman deployed two new servers running a new and improved version of the Bits caching server in Amsterdam, to prepare for the deployment in Ashburn.

Media Storage [?] — During the New Orleans hackathon, developers agreed to create an abstraction layer (FileBackend) which will allow MediaWiki users to use SwiftMedia to store media files, or another filesystem of their choice. They also agreed to re-architect the FileRepo module in Mediawiki.

HTTPS — HTTPS support has been enabled on all Wikimedia sites; mobile support is still to come. At some point in the future, we'll be switching the log-in link to default to logging in via https.

Testing environment

Virtualization test cluster [?] — Wikimedia Labs has been launched. We are starting to install instances to duplicate our production environment. We are also bringing up some instances for non-MediaWiki related work. At the NOLA hackathon, we worked with a team from Canonical to give the juju project access to our environment. A new project from Wikimedia Deutschland, Wikidata, announced they'd be working in the Labs environment as well.

Backups and data archives

Data Dumps [?] — This month's dump for the English Wikipedia started later in the month (on October 11), since we were waiting for the production roll-out of MediaWiki 1.18. The run completed without issues, and all other servers are also using 1.18 without problems. We set up our first mirror site, hosted by the C3SL at Universidade Federal do Paraná in Brazil, whom we thank for their support. The mirror provides a copy of the last 5 successful dumps of each project and is updated as new dumps complete.

Features Engineering

Editing tools

Visual editor — Trevor Parscal worked on a new model tree (and the conversion from the linear model) and updated the software design documentation. Inez Korczynski worked on front-end functionality: dialogs, selecting text, mouse actions, scrolling, keyboard shortcuts, etc. Roan Kattouw focused on algorithms that insert ranges of data into a document, and Neil Kandalgaonkar on the removal of data. On the parser side, a basic parser using PEG is in place, which produces an intermediate JSON object tree; Brion Vibber is still working on markup support (mixed HTML and Wikitext). Gabriel Wicke joined the team and started to work on the PEG parser (read more).
Internationalization and localization tools — Siebrand Mazeland and the rest of the team continued to plan their work using Mingle (credentials: guest/guest). Santhosh Thottingal added new features, and support for more languages, to the WebFonts extension, as well as new keyboard mappings and languages to the Narayam input method extension. Gerard Meijssen started to build language support teams focusing on localization and internationalization. A major achievement was the deployment of the Translate extension to meta-wiki, by Niklas Laxström and Sam Reed; it will be used to translate content pages, e.g. for the 2011 fundraiser. The localization team ended its first sprint as a team, and presented a development showcase.

Participation and editor retention

Article feedback — Work has started on Version 5 of this feature, which will focus on reader engagement, for example by offering new feedback forms and free-text comment fields; a comment moderation system is also planned. This project is led by contractor Fabrice Florin, and will be implemented by an external contracting company, OmniTI.
LiquidThreads 3.0 — Andrew Garrett started working on a conversion script to facilitate the transition from the old schema to the new one.
Feedback Dashboard — The Feedback dashboard was deployed to the English Wikipedia, showing information provided by MoodBar. It is primarily read-only: it isn't possible to reply to comments. Andrew Garrett implemented the ability for administrators to hide comments which violate terms of use.

Multimedia Tools

UploadWizard — Ian Baker implemented multi-file selection. Neil Kandalgaonkar did research on how licenses are used on Commons, added custom wikitext licenses with various safeguards to ensure this wouldn't result in more poorly licensed files, and implemented wording changes on the advice of Wikimedia general counsel. Ben Hartshorne implemented a simple way to add geocoding templates. Jeroen De Dauw also worked on this feature.

MediaWiki infrastructure

ResourceLoader — Roan Kattouw and Timo Tijhof set up a prototype wiki farm to test the gadget repository, and continued to add features like gadget creation in the gadget manager.

Media Labs

Multimedia — Ian Baker and Neil Kandalgaonkar continued to review Michael Dale's code to prepare it for deployment. Ben Hartshorne has set up the initial development environment for Engineering to have a platform on which to test and continue development of the SwiftMedia extension. Ben and Mark Bergsma continue to do performance testing on Swift prior to using it in production. Aaron Schulz started to refactor the file backend code, a requirement to using SwiftMedia.

Mobile

Mobile Research [?] — Mani Pande and Parul Vora conducted user experience research in the US in three cities (San Francisco, Chicago and Dallas) assisted by UX firm AnswerLab. They met with current and potential mobile readers, as well as editors. They are also finalizing the report on their field research in India and Brazil.

MobileFrontend — The production version of our new Wikimedia mobile platform, launched in September, is no longer considered to be beta. Users can opt-in to the beta program to test new, pre-release features and provide feedback; future features include search suggestions and interwiki links. Phil Chang also set up a mobile features brainstorm page, and started to organize them. In related news, the Wikimedia Foundation is looking for mobile operator partners for its Wikipedia Zero program.

Android Wikipedia App [?] — We partnered with nitobi to create the first official Wikipedia app for Android devices. The app is based on the PhoneGap framework and should be added to the Android market soon. The team is still looking for developers.

Special projects

Fundraising support

2011 Fundraiser — October was a very busy month for fundraising engineering, as the annual fundraiser is about to start. The biggest achievement was the rewrite and deployment of the credit card processing software, the DonationInterface extension. It now supports GlobalCollect, a new payment processor, in addition to the existing ones. Work is underway to add support for non-credit card payments. Another major feature of the extension is the ability to easily create new forms simply by inserting template tags into HTML markup; it also leverages MediaWiki's message translation system using translatewiki.net. Changes were made to the LandingCheck extension to determine whether to send users to donate.wikimedia.org (used by the Wikimedia Foundation) or wikimediafoundation.org (used by some Wikimedia chapters). Donators can now opt out from e-mails directly from the thank-you message they receive after their donation. Last, from an operations perspective, we increased the monitoring capacity of the servers in the payment cluster, as well as the redundancy of systems; we're currently working to implement automatic failover between our data centers.

Offline

Kiwix UX initiative [?] — The last critical bugs in version 0.9 of Kiwix (especially on Windows) were fixed, and the release candidate cycle will soon begin. A Kiwix activity was released in alpha for Sugar, a learning platform for children. The first two prototypes of the Kiwix-plug have been built; the Kiwix-plug is a standalone wifi hotspot providing offline Wikipedia content, based on a plug computer.

Platform Engineering

MediaWiki Core

MediaWiki 1.18 — The deployment of MediaWiki 1.18 was successfully completed on all Wikimedia wikis. Developers are now fixing the remaining bugs before releasing 1.18 for third parties. The first beta is expected in early November.
Code review management — The focus on 1.18, as well as events and conferences, have allowed the backlog of unreviewed commits to increase in October (see chart). There are currently about 350 unreviewed revisions in trunk/phase3/, which means the deployment of MediaWiki 1.19 might not happen until January 2012.
Shell requests — Sam Reed continued to follow up on and fulfill shell requests.
Continuous integration — Chad Horohoe worked with the operations team to finalize the setup of the testing server, now online at https://integration.wikimedia.org. It is currently running Jenkins (PHPUnit), and TestSwarm should be added soon. Antoine Musso will be leading this project going forward.
Git conversion — Chad Horohoe started to prepare for the migration to git; he wrote a script to compare LDAP accounts to the information stored in Code review, and prepared for conversion tests from the current Subversion repository. An important step in the conversion is having names and email addresses for all committers, since neither Git nor Gerrit function well without them. As of this writing, 26 committers do not have any contact information, and 25 have a name but no e-mail address. Please help by filling in the names of committers you know.

Wikimedia analytics

Wikimedia Report Card 2.0 — Erik Zachte discovered inconsistencies in the report card numbers, which were investigated and attributed to packet loss of up to 25%. This has now been fixed; further steps to prevent similar issues include adding monitoring to the process, and versioning the configuration files.

Technical Liaison; Developer Relations

Bug management — Mark Hershberger continued to hold themed bug triage sessions; in October, those focused on Fundraiser engineering and the new Wikipedia Android app. Two ad-hoc triages, on shell requests and 1.18 regressions, happened during the New Orleans hackathon. Mark also worked with members of the "MediaWiki Core" group to better identify "highest priority" bugs in Bugzilla and continued to prioritize and respond to new bug reports.
Summer of Code 2011 — Sumana Harihareswara, Neil Kandalgaonkar and Max Semenik attended the GSoC mentor summit in Mountain View, California. Students Salvatore Ingala and Kevin Brown made progress on integrating their summers' work into MediaWiki trunk, during the New Orleans hackathon; much of the GSoC students' work remains to be integrated into MediaWiki core or extensions.
Engineering project documentation — Guillaume Paumier performed continuous maintenance of engineering project pages, and assembled this report.
Volunteer coordination and outreach — Sumana Harihareswara continued to go through the backlog of commit access requests, and participated in a discussion on making it easier to get Subversion commit access. In October, eleven people got commit access, of which six were volunteers and five were WMF staff or contractors. She began to organize a triage of database-related bugs to benefit MediaWiki developers and administrators who use RDBMSes other than MySQL, and encouraged volunteers who are interested in leading hackathons, teaching classes, or attended conferences to talk about MediaWiki. She also followed up with volunteers and potential new developers, notably from the New Orleans hackathon and Google Summer of Code mentors' summit.
MediaWiki architecture document — Guillaume Paumier finished the write-up of the document, based on the input provided by developers, the existing documentation on mediawiki.org, the auto-generated documentation from doxygen, and deep dives into the actual code. The MediaWiki community reviewed the document, which was submitted to the book's editors. This project is considered to be mostly completed. Further work will include minor polishing and follow-up on the document and its publication, and integration of the content into the relevant pages on mediawiki.org.

Future

The engineering management team continues to update the Software 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.