in MediaWiki

tagged mediawiki, notifications

The future of Echo

Echo (aka "Notifications") is the MediaWiki extension that provides a notifications framework for other features to use, as well as some "core" notification types. It's had a tumultuous history ever since it was deployed to Wikimedia wikis in 2013. To figure out where Echo should go, I think we have to look at its history first.

The initial deployment was pretty rough. The OBOD was gone, but no replacement. Standard APIs like hasmsg=1 didn't work either. The development team (WMF's "E2" team) iterated on those, improved the API integration and added a newer and less obtrusive new messages indicator (while also designing my personal favorite, nerdalert.js). So the time came for the deployment to be expanded to all Wikimedia wikis, at which point nearly the entire development team switched over to a different project (Flow), leaving one engineer to supervise the rollout over all projects. Umm, what???

So we ended up with bugs like "Mention notifications don't work if the sender's signature contains localized namespaces" taking nearly 5 full months to fix. That sucked.

Anyways, Echo was mostly dormant until mid-2014, when the "Core features" (previously E2) team made changes to Echo to support having Flow notifications go in a separate pane in the flyout. Except Flow was barely used at this point, so no one noticed outside of mediawiki.org really. (I was technically on the core features team at this point, but mostly doing Flow API things IIRC). But after a month or two most of the development moved back to Flow.

And then finally after the great engineering re-org of 2015, the Collaboration team (formerly Core features, but you already guessed that ;-)) also started looking at Echo seriously, and starting to fix some of the issues that had piled up over the years, including splitting the alerts and messages flyouts properly, giving user talk messages more prominence, and eventually embarking upon cross-wiki notifications.

Despite barely getting any attention from developers (until now really), Echo remains the most popular and really only successful product to come out of the E2/Core features/Collaboration team. Why?

The most useful feature of Echo is definitely the mention notifications which allow you to "ping" other users. So instead of people having to watchlist giant pages and look through history to see if anyone responded to the one thread they want to follow, they can wait for the notification that someone pinged them while responding. The once widely used {{talkback}} template is now deprecated in favor of these notifications. And for most users, this functionality is good enough. Watchlists really haven't seen any major changes in the past few years (again, until the past month, but that's another story), so something else that can do the job was welcomed by users.

So now we're in the state where we have two overlapping, but not exactly the same features: Notifications and Watchlists. Gryllida has written up an RfC titled "Need to merge Notifications and Watchlist or lack thereof", discussing some of the similarities and differences. Over the next few months I'd like to flesh out the RfC a bit more and work on a solid proposal.

I also wrote an RfC yesterday titled "Notifications in core", which discusses merging parts of the Echo extension into MediaWiki core. I think this is crucial in improving the usability of notifications from both a user and developer perspective, as well as improving the architecture by requiring less hacks. And it can probably be done before the reconciliation of notifications and watchlists.

So, that's where I think Echo should go in the next year or two. I probably won't have time to actually work on this, so we'll see!

in MediaWiki

tagged notifications, wikimedia

Why I am still here.

This is a follow up to my probably depressing blog post on Friday, entitled "Why am I still here?" I think I figured out the answer, it's a lot of different things!

So now that you're here, lets talk about something happy: cross-wiki notifications!

In 2013, notifications were introduced and collected events that users wanted to know about in a flyout. The initial response was that it looked like Facebook (it really did), and anger (well I'm not sure what the right emotion was) that the "Orange Bar of Death" was gone. But the longer term impact has been a significant change to how on-wiki discussions work. One of the new features was that users can get "pinged" when someone links their username in a discussion comment. These days usage of templates like {{talkback}} is extremely rare, and entirely replaced with pinging.

So, cross-wiki notifications. This has been a long requested feature request, and I had actually written an RfC about implementing it long before I was on the team tasked with implementing it (well, technically I was on the E2 team at that time, but history is confusing). Of course, I would be remiss to mentoin that most editors already have a cross-wiki notification system through their email inbox. Regardless, some people are so excited they didn't even read the announcement. ;-)

However, there were some limitations in the architecture that made expanding it for cross-wiki usage difficult. The most significant was that the formatting system was designed for specific types of notifications, hard to extend, and most importantly, difficult for developers to understand. We probably could have hacked our way around it, but the extensability was going to be a serious problem. And if we were going to touch some code, we should leave it better than we found it. :)

So, what to replace it with? I came up with a presentation model system that every notification type would implement. This made it much easier to follow instead of the old data-driven style of formatters. We also built an API (sorry, logged in users only) around it, so the frontend code can handle all of the display, and not have to parse out HTML to figure out the notification icon (yes, it used to do that :().

So now...going cross-wiki. We effectively had two approaches we could do, the first being created a central database table that all notifications went into, and then having each wiki read from it. This provided some logistical problems like checking revdel status, page existence, and local message overrides. The other approach was to make API requests to other wikis, and then merge the output into the flyout. We ended up going with the second one (cross-wiki API requests) after seeing crosswatch in action, which used the same strategy for its implementation of cross-wiki notifications.

We created a database tracking table which contains the number of unread alerts and messages you have on each wiki, so it would be known on what wikis to query notifications for. We also realized we could make these requests client-side (using CORS) for now, and move them server-side later.

And that's pretty much it. There were also some significant frontend changes, like converting it all to use OOjs UI, but I didn't really work on those, and can't speak much about them.

So, go to Special:BetaFeatures on your favorite wiki, check "enhanced notifications", and let us know what you think! And if it's not on your wiki yet, nag someone in #wikimedia-collaboration on freenode!

Update: The list of wikis where it is available are listed on MediaWiki.org, and it should hit all other wikis by the end of March.

in MediaWiki

tagged wikimedia

Why am I still here?

Honestly, I'm not sure. I'm not working at the WMF for the money. And it's not because my job is fun - it stopped being fun a few months ago. I'm mostly still here because I don't really have anything else to do right now. I've been doing this for nearly 3 years now...and even though I have other plans lined up in a few months, I'm kind of a lame duck right now.

I'm embarrassed at everything that's gone on. I don't want to tell my friends that the WMF is all screwed up and not going to get better. I don't want to have to explain how after I praised how all of my work is public and transparent, that our organization no longer is.

The whole Arnnon situation never really sank in for me. It was mostly surreal, with me thinking that there was no way they'd actually let him stay on the board. All the right people spoke up, signed the petition, and then he was removed. The process worked! Not at all. The excuse of not knowing about it because they didn't use "google.com" is ludicrous, that I don't even know what to say about it.

It greatly frustrates me, and upsets me that my friends are in danger of having to leave the country if they lose their jobs, and can't speak up because of that fear.

Honestly, I don't understand why the current leadership hasn't left yet. Why would you want to work at a place where 93% of your employees don't believe you're doing a good job, and others have called you a liar (with proof to back it up) to your face, in front of the entire staff? I don't know everything that's going on right now, but we're sick right now and desparately need to move on.

I love, and will always love Wikimedia, but I can't say the same about the current state of the Wikimedia Foundation. I've been around for nearly nine years now (nearly half my life), and it feels like that world is slowly crumbling away and I'm powerless to stop it.

And that's all my incoherent rambling for tonight.

And now for something less depressing and totally awesome, I mostly-finished the first version of the WikimediaPageViewInfo extension, you can see a demo at http://bf-wmpageview.wmflabs.org/wiki/Taylor_Swift?action=info. It's currently going through a security review, and myself and Addshore are going to work on getting it deployed!