The last quarter of 2018 has been a quite hectic one, and every time I had some spare time after the Web Engines Hackfest the prospective of sitting down to write some thoughts about it seemed dreadful. Christmas went by already —two full weeks of holidays, practically without touching a computer— and suddenly I found myself booking tickets to this year’s FOSDEM and it just hit me: it is about time to get back blogging!

FOSDEM

There is not much that I would want to add about FOSDEM, an event which I have attended a number of times before (and some others about which I have not even blogged). This is an event I always look forward to, and the one single reason that keeps me coming back is recharging my batteries.

This may seem contradictory because the event includes hundreds of talks and workshops tucked in just two days. Don’t get me wrong, the event is physically tiresome, but there are always tons of new and exciting topics to learn about and many Free/Libre Software communities being represented, which means that there is a contagious vibe of enthusiasm. This makes me go back home with the will to do more.

Last but not least, FOSDEM is one of these rare events in which I get to meet many people who are dear to me — in some cases spontaneously, even without knowing we all would be attending. See you in Brussels!

Web Engines Hackfest

Like on previous years, the Web Engines Hackfest has been hosted by Igalia, in the lovely city of A Coruña. Every year the number of participants has been increasing, and we hit the mark of 70 people in the 2018 edition.

Are We GTK+4 Yet?

This time I was looking forward to figuring out how to bring WebKitGTK+ into the future, and in particular to GTK+4. We had a productive discussion with Benjamin Otte which helped a great deal to understand how the GTK+ scene graph works, and how to approach the migration to the new version of the toolkit in an incremental way. And he happens to be a fan of Factorio, too!

In its current incarnation the WebKitWebView widget needs to use Cairo as the final step to draw its contents, because that is how widgets work, while widgets in GTK+4 populate nodes of a scene graph with the contents they need to display. The “good” news is that it is possible to populate a render node using a Cairo surface, which will allow us to keep the current painting code. While it would be more optimal to avoid Cairo altogether and let WebKit paint using the GPU on textures that the scene graph would consume directly, I expect this to make the initial bringup more approachable, and allow building WebKitGTK+ both for GTK+3 and GTK+4 from the same code base. There will be room for improvements, but at least we expect performance to be on par with the current WebKitGTK+ releases running on GTK+3.

While not needing to modify our existing rendering pipeline should help, and probably just having the WebKitWebView display something on GTK+4 should not take that much effort, the migration will still be a major undertaking involving some major changes like switching input event handling to use GtkEventController, and it will not be precisely a walk in the park.

As of this writing, we are not (yet) actively working on supporting GTK+4, but rest assured that it will eventually happen. There are other ideas we have on the table to provide a working Web content rendering widget for GTK+4, but that will a the topic for another day.

The MSE Rush

At some point people decided that it would be a good idea to allow Web content to play videos, and thus the <video> and <audio> tags were born. All was good and swell until people wanted playback to adapt to different types of network connections and multiple screen resolutions (phones, tablets, cathode ray tubes, cinema projectors…). The “solution” is to serve video and audio in multiple small chunks of varying qualities, which are then chosen, downloaded, and stitched together while the content is being played back. Sci-fi? No: Media Source Extensions.

A few days before the hackfest it came to our attention that a popular video site stopped working with WebKitGTK+ and WPE WebKit. The culprit: The site started requiring MSE in some cases, without supporting a fallback anymore, our MSE implementation was disabled by default, and when enabled it showed a number of bugs which made it hardly possible to watch an entire video in one go.

During many the Web Engines Hackfest a few of us worked tirelessly, sometimes into the wee hours, to make MSE work well. We managed to crank out no less than two WebKitGTK+ releases (and one for WPE WebKit) which fixed most of the rough edges, making it possible to have MSE enabled and working.

And What Else?

To be completely honest, shipping the releases with a working MSE implementation made the hackfest pass in a blur and I cannot remember much else other than having a great time meeting everybody, and having many fascinating conversations — often around a table sharing good food. And that is already good motivation to attend again next year 😉