Editor retention: Editing tools
In July, the team working on VisualEditor converged the design for mobile and desktop, made it possible to see and edit HTML comments, improved access to re-using citations, and fixed over 120 bugs and tickets.
The new design, with controls focussed at the top of each window in consistent positions, was made possible due to the significant progress made in cross-platform support in the UI library, which now provides responsively-sized windows that can work on desktop, tablet and phone with the same code. HTML comments are occasionally used on a few articles to alert editors to contentious or problematic issues without disrupting articles as they are read, so making them prominently visible avoids editors accidentally stepping over expected limits. Re-using citations is now provided with its simple dialog available in the toolbar so that it is easier for users to find.
Other improvements include an array of performance fixes targeted at helping mobile users especially, fixing a number of minor instances where VisualEditor would corrupt the page, and installing better monitoring of corruptions if they occur, and better support for right-to-left languages, displaying icons with the right orientation based on context.
The mobile version of VisualEditor, currently available for beta testers, moved towards stable release, fixing a number of bugs and editing issues and improving loading performance. Our work to support languages made some significant gains, nearing the completion of a major task to support IME users, and the work to support Internet Explorer uncovered some more issues as well as fixes. The deployed version of the code was updated five times in the regular release cycle (1.24-wmf12, 1.24-wmf13, 1.24-wmf14, 1.24-wmf15 and 1.24-wmf16).
In wider news, the team expanded its scope to cover all MediaWiki editing tools as well, as the new Editing Team (covered below).
In July, the newly re-named and re-scoped Editing Team was formed from the VisualEditor Team. We are responsible for extending and improving the editing tools used at Wikimedia – primarily VisualEditor and maintenance for WikiEditor. We exist to support new and existing editors alike; our current work is mostly on desktop, and we are working with Mobile to take responsibility for all editing across desktop, tablet and phone platforms, spanning approximately 50 different areas of MediaWiki and extensions related to editing. We will continue to report progress on VisualEditor separately.
The biggest Editing change this month was in the Cite extension (for footnotes) – this now automatically shows a references list at the end of the page if you forget to put in a
<references /> tag, instead of displaying an ugly error message. The Math extension (for formulæ) was improved with more rigorous error handling and LaTeX formula checking, as part of the long-term volunteer-led work to introduce MathML-based display and editing. The TemplateData GUI editor was deployed to a further six wikis – the English, French, Italian, Russian, Finnish and Dutch Wikipedias.
A lot of work was done on libraries and infrastructure for the Editing Team and others. The OOjs UI library was extensively modified to bring in a new window management system for comprehensive combined desktop, tablet and phone support, as well as other updates to improve Internet Explorer compatibility and accessibility of controls. In the next few months the team will continue working on OOUI to support other teams’ needs and implement a consistent look-and-feel in collaboration with the Design team. The OOjs library was updated to fix a minor bug, with a new version (v1.0.11) released and pushed downstream into MediaWiki, VisualEditor and OOjs UI. The ResourceLoader framework was extended to allow skins to set the “skinStyles” property themselves, rather than rely on faux dependencies, as part of wider efforts led jointly by a volunteer and a team member to improve MediaWiki’s skin support.
In July, the Parsoid team continued with ongoing bug fixes and bi-weekly deployments.
With an eye towards supporting Parsoid-driven page views, the Parsoid team strategized on addressing Cite extension rendering differences that arise from site-messages based customizations and is considering a pure CSS-based solution for addressing the common use cases. We also finished work developing the test setup for doing mass visual diff tests between PHP parser rendering and Parsoid rendering. It was tested locally and we started preparations for deploying that on our test servers. This will go live end-July or early-August.
The GSoC 2014 LintTrap project continued to make good progress. We had productive conversations with Project WikiCheck about integrating LintTrap with WikiCheck in a couple different ways. We hope to develop this further over the coming months.
Overall, this was also a month of reduced activity with Gabriel now officially full time in the Services team and Scott focused on the PDF service deployment that went live a couple days ago. The full team is also spending a week at a off-site meeting working and spending time together in person prior to Wikimania in London.
Services and REST API
The brand new Services group (currently Matt Walker and Gabriel Wicke) started July with two main projects:
- PDF render service deployment
- Design and prototyping work on the storage service and REST API
The PDF render service is now deployed in production, and can be selected as a render backend in Special:Book. The renderer does not work perfectly on all pages yet, but the hope is that this will soon be fixed in collaboration with the other primary author of this service, C. Scott Ananian.
Prototyping work on the storage service and REST API is progressing well. The storage service now has early support for bucket creation and multiple bucket types. We decided to configure the storage service as a backend for the REST API server. This means that all requests will be sent to the REST API, which will then route them to the appropriate storage service without network overhead. This design lets us keep the storage service buckets very general by adding entry point specific logic in front-end handlers. The interface is still well-defined in terms of HTTP requests, so it remains straightforward to run the storage service as a separate process. We refined the bucket design to allow us to add features very similar to Amazon DynamoDB in a future iteration. There is also an early design for light-weight HTTP transaction support.
Matt Walker is sadly leaving the Foundation by the end of this month to follow his passion of building flying cars. This means that we currently have three positions open in the service group, which we hope to start filling soon.
In July, the Flow team built the ability for users to subscribe to individual Flow discussions, instead of following an entire page of conversations. Subscribing to an individual thread is automatic for users who create or reply to the thread, and users can choose to subscribe (or unsubscribe) by clicking a star icon in the conversation’s header box. Users who are subscribed to a thread receive notifications about any replies or activity in that thread. To support the new subscription/notification system, the team created a new namespace, Topic, which is the new “permalink” URL for discussion threads; when a user clicks on a notification, the target link will be the Topic page, with the new messages highlighted with a color. The team is currently building a new read/unread state for Flow notifications, to help users keep track of the active discussion topics that they’re subscribed to.
In July, the Growth team completed its second round of A/B testing of signup invitations for anonymous editors on English Wikipedia, including data analysis. The team also built the first API and interface prototypes for task recommendations. This new system, first aimed at brand new editors, makes suggestions based on a user’s previous edits.
Following on from the successful launch to Android, the Mobile Apps team released the new native Wikipedia app to iOS on July 31. The app is the iOS counterpart to the Android app, with many of the same features such as editing, saving pages for offline reading, and browsing history. The iOS app also contains an onboarding screen that is shown the first time the app is launched, asking users to sign up, a feature which was also launched on Android this month (see below).
On Android this month we released to production accessibility and styling features which were requested by our users, such as a night mode for reading in the dark and a font size selector. We also released an onboarding screen that asks users to sign up.
Our plan for next month is to get user feedback from Wikimania, wrap up our styling fixes, and begin work on an onboarding screen the first time that someone taps edit.
Mobile web projects
This month, the team continued to focus on wrapping up the collaboration with the Editing team to bring VisualEditor to tablet users on the mobile site. We also began working to design and prototype our first new Wikidata contribution stream, which we will build and test with users on the beta site in the coming month.
During the last month, the team worked on software architecture features that allow for expansion of the Wikipedia Zero footprint on partner networks and that get users to content faster with support for lowered cache fragmentation on Varnish caches. Whereas the previous system supported one-size-fits-all configuration for heterogeneous partner networks, inhibiting some zero-rated access, the new system supports multiple configurations for disparate IP addresses and connection profiles per operator. Additionally, lightweight script and GIF-ified Wikipedia Zero banner support has been added and is being tested; in time this should drastically reduce Varnish cache fragmentation, making pages be served faster and reducing Varnish server load. A faster landing page was introduced for “zerodot” (zero.wikipedia.org, legacy text-only experience) landing pages when operators have multiple popular languages in their geography. Work on compression proxy traffic analysis for header enrichment conformance with the official Wikipedia Zero configurations was also performed after more diagnostic logging code was added to the system. Finally, watchlist thumbnails, although low bandwidth, were removed from the zerodot user experience, as was the higher bandwidth MediaViewer feature for zerodot; mdot will have these features, though.
In side project work, the team spent time on API continuation queries, Android IP editing notices, Amazon Kindle and other non-Google Play distribution, and Google Play reviews (now that the Android launch dust has settled, mobile apps product management will be triaging the reviews). In partnerships work, the team met with Mozilla to talk about future plans for the Firefox OS HTML5 app (e.g., repurposing the existing mobile website, but without any feature reduction) and how Wikimedia search might be further integrated into Firefox OS, and also spoke with Canonical about how Wikipedia might be better integrated into the forthcoming Ubuntu Phone OS.
Routine pre- and post-launch configuration changes were made to support operator zero-rating, with routine technical assistance provided to operators and the partner management team to help add zero-rating and address anomalies. The team also continued its search for a third Partners engineering teammate.
Wikipedia Zero (partnerships)
- We served an estimated 68 million free page views in July through Wikipedia Zero. We continue to bring new partners into the program, though none launched in July. Adele Vrana met with prospective partners and local Wikimedians in Brazil. We published our operating principles to increase transparency.
The Beta cluster
is running HHVM. The latest MediaWiki-Vagrant
use HHVM by default.
Admin tools development
Most admin tools resources are currently diverted towards SUL finalisation
, which will greatly help in reducing the admin tools backlog. July saw the deployment of the global rename tool (bug 14862
), and core fixes including the creation of the “viewsuppressed” userright (bug 20476
Our deployment of CirrusSearch to larger wikis as the primary search back-end turned out to be too ambitious. After encountering performance issues, we rolled back this change. We are now addressing the root of the problem, by getting more servers (nearly doubling the cluster size) and putting together more optimizations to the portion of Cirrus that fell over (working set). If everything goes as planned, it’ll be reduced by about 80%, by reducing indexing performance in return of search performance. These optimizations will slightly change result relevance; please let us know if you notice any issues.
Most work was spent on SUL Finalization tasks. Phpunit and browser tests were added for CentralAuth, global rename was deployed, and lots of small fixes were made to CentralAuth to clean up user accounts in preparation for finalization.
In July, the SUL finalisation team began work on completing the necessary feature work to support the SUL finalisation.
To help users with local-only accounts that are going to be forcibly renamed due to the SUL finalisation, the team is working on a form that lets those users request a rename. These requests will be forwarded onto the stewards to handle. The SUL team is currently in consultation with the stewards about how they would like this tool to work. When this consultation is wrapped up, the team will begin design and implementation.
To help users get globally renamed without having to request renames on potentially hundreds of wikis, the team implemented and deployed GlobalRenameUser, a tool which renames users globally. As the tool is designed to work post-finalisation, it only performs renames where the current name is global, and the requested name is totally untaken (no global account and no local accounts exist with that name).
To help users who get renamed by the finalisation and, despite our best efforts to reach out to them, did not get the chance to request a rename before the finalisation, the team is working on a feature to let users log in with their old credentials. The feature will display an interstitial when they log in, informing them that they logged in with old credentials and that they need to use new ones. We are also considering a persistent banner for those users, so that they definitely know they need to use their new credentials. An early beta version of this feature is complete, and now needs design and product refinements to be completed.
To help users who get renamed by the finalisation and, as a result, have several accounts that were previously local-only turned into separate global accounts, the team is working on a tool to merge global accounts. We chose to merge accounts as it was the easiest way to satisfy the use case without causing further local-global account clashes that would cause us to have to perform a second finalisation. The tool is in its preliminary stages.
The team also globalised some accounts that were not globalised but had no clashes. These accounts were either created in this local-only form due to bugs, or are accounts from before CentralAuth was deployed where the user never globalised. As these accounts had no clashes, there were no repercussions to globalising these accounts, so we did this immediately.
At present, no date has been chosen for the finalisation. The team plans to have the necessary engineering work done by the end of the quarter (end of September 2014), and have a date chosen by then.
Next month the team plans to continue work on these features.
Security auditing and response
MediaWiki 1.23.2 was released, fixing 3 security bugs. Security reviews were made for BounceHandler and Petition extensions, and the password API was merged.
This month, the Release and QA Team became the Release Engineering Team, mostly reflecting the transition of this team from being made up of members of other distinct teams to that of a coherent self-contained (mostly) team. This will, hopefully, allow better coordination of “Release” and “QA” things (broadly spreaking).
A lot of progress was made on making Phabricator suitable as a task/bug tracking system for Wikimedia projects. You can see the work to be sorted and completed at this workboard.
The Beta Cluster now runs with HHVM, bringing us much closer to full HHVM deployment. In addition, the Language Team deployed the new Content translation system on the Beta Cluster with the help of the Release Engineering team.
The second round of public RFP for third-party MediaWiki release management was conducted and concluded.
We now no longer use the third-party Cloudbees service for any of our Jenkins jobs and run all jobs locally. This will enable us to better diagnose issues with our build process, especially as it pertains to our browser tests (which still mostly run on SauceLabs).
This month, the QA team finished two significant achievements: after porting all the remaining browser tests from the
browsertests repository to the repositories of the extensions being tested in June, as well as porting a significant set of tests to MediaWiki core itself, we completely retired the Jenkins instance running on a third-party host in favor of running test builds from the Wikimedia Jenkins instance, and we deleted the
/qa/browsertests code repository. These moves are the result of more than two years of work. In addition, we have added more functions to the API wrapper used by browser tests, improved support for testing in Vagrant virtual machines, added new Jenkins builds for extensions, and improved the function of the beta labs test environments by preventing database locks and stopping users from being logged out by accident.
The browser tests are now all integrated with builds on the Wikimedia Jenkins host. We added browser tests for MediaWiki core that will validate the correctness of a MediaWiki installation regardless of language, or of what extensions may or may not exist on the wiki, so that the tests may be packaged with the distribution of MediaWiki itself and used on arbitrary wikis. We saw a lot of browser test activity for Flow development, and we are preparing to support even more extensions and features in the very near future.
In July, the multimedia team reviewed more feedback about Media Viewer, from three separate Requests for Comments on the English and German Wikipedias, as well as on Wikimedia Commons. Based on this community feedback, the team worked to make the tool more useful for readers, while addressing editor concerns. We are now considering a new ‘minimal design’, which would include: a much more visible link to the File: page; an even easier way to disable the tool; a caption or description right below the image; removing additional metadata below the image, directing users to the File: page instead.
As described in our improvements plan, these new features are being prototyped and will be carefully tested with target users in August, so we can validate their effectiveness before developing and deploying them in September. You can see some of our thinking in this presentation.
This month, we continued to work on the Structured Data project with the Wikidata team and many community members, to implement machine-readable data on Wikimedia Commons. We prepared to host a range on online and in-person discussions to plan this project with our communities, and aim to develop our first experiments in October, based on their recommendations. We also continued a major code refactoring for the UploadWizard, as well as fixed a number of bugs for some of our other multimedia tools.
Last but not least, we prepared seven different multimedia roundtables and presentations for Wikimania 2014, which we will report on in more depth in August. For now, you can keep up with our work by joining the multimedia mailing list.
All Google Summer of Code and FOSS Outreach Program for Women projects continued their development toward a successful end. For details, check the reports:
Guillaume Paumier collaborated with authors of the Education newsletter to set it up for multilingual delivery, using a script similar to the one used for Tech News. He also wrote a detailed how-to to accompany the script for people who want to send a multilingual message across wikis. In preparation for the Wikimania session about Tech News, he updated the readability and subscribers metrics. He also continued to provide ongoing communications support for the engineering staff, and to prepare and distribute Tech News every week.
Volunteer coordination and outreach
We focused on the preparation of the Wikimania Hackathon
, encouraging all registered participants to propose topics and sign up to interesting sessions. We also organized a Q&A session
with potential organizers of the Wikimedia Hackathon 2015. We organized two Tech Talks: Hadoop and Beyond. An overview of Analytics infrastructure
and HHVM in production: what that means for Wikimedia developers
. More activities hosted in July can be found at Project:Calendar/2014/07
Architecture and Requests for comment process
Wikimetrics can now generate vital sign metrics for every project daily. Rolling Monthly Active Editor metric has been implemented; the reports are in JSON format, in a logical path hosted on a file server and downloadable. The team also worked on backfilling data for the daily reports on Newly Registered and Rolling Active Editor, and numerous optimizations to backfill the data quickly.
New nodes were added to the cluster this month and all machines were upgraded to run CDH5. The team decided not to preserve any data on the cluster during the upgrade and started fresh. The team hosted a Tech Talk on our Hadoop installation (see video
). Duplicate monitoring has also been implemented in Hadoop to monitor the incoming Varnish logs.
Editor Engagement Vital Signs
The culmination of our efforts this month can be visualized in a prototype built for Wikimania
. This was made possible thanks to many back-end enhancements (optimizations) to Wikimetrics, along with research and selection of the optimal technologies to implement the stack to display a dashboard
EventLogging monitoring is now in graphite, and we can see which schemas cause spikes in traffic (example
Research and Data
This month, we completed the documentation for the Active Editor Model, a set of metrics for observing sub-population trends and setting product team goals. We also engaged in further work on the new pageviews definition. An interim solution for Limited-duration Unique Client Identifiers (LUCIDs) was also developed and passed to the Analytics Engineering team for review.
We analyzed trends in mobile readership and contributions, with a particular focus on the tablet switchover and the release of the native Android app. We found that in the first half of 2014, mobile surpassed desktop in the rate at which new registered users become first-time editors and first-time active editors in many major projects, including the English Wikipedia. An update on mobile trends will be presented at the upcoming Monthly Metrics meeting on July 31.
Development of a standardised toolkit for geolocation, user agent parsing and accessing pageviews data was completed.
We supported the multimedia team in developing a research study to objectively measure the preference of Wikipedia editor and readers.
We hosted the July research showcase with a presentation by Aaron Halfaker of 4 Python libraries for data analysis, and a guest talk by Center for Civic Media’s Nathan Matias on the use of open data to increase the diversity of collaboratively created content.
We prepared 8 presentations that we will be giving or co-presenting next week at Wikimania in London. We also organized the next WikiResearch hackathon that will be jointly hosted in London (UK) (during the pre-conference Wikimania Hackathon) and in Philadelphia (USA) on August 6-7, 2014.
We filled the fundraising research analyst position: the new member of the Research & Data team will join us in September and we’ll post an announcement on the lists shortly before his start date.
Lastly, we gave presentations on current research at the Wikimedia Foundation at the Institute for Scientific Interchange (Turin) and at the DesignDensity lab (Milan).