What programming/technical projects have you been working on?

I'm working on sandboxing content processes in Firefox for Android.

When it comes to web browsers, adding the code to actually do the sandboxing is easy; the hard part is adapting the rest of the browser's code to be aware of it.

Essentially, there are many components in the browser that were written for a pre-sandboxing universe. These components expect to be able to talk directly to the OS, and potentially to other processes, without hindrance. As we make the sandbox stricter, we need to modify those components to work differently. Often times this involves changing the affected components so that, when running as sandboxed content, they communicate with their counterparts in the parent process to do work on their behalf. In Mozilla parlance, we refer to the procedure of modifying these components as "remoting."

Currently working on a graphical Gemini client because apparently I didn't learn my lesson last time.

https://sr.ht/~admicos/moonlander
Obligatory Screenshot

At the time of writing it's pretty "half-baked" and not daily-driver ready just yet, but it's definitely more promising than my previous attempt. The main difference between them being that I am not trying any custom drawing magic anymore, and instead (ab)using some GTK features to achieve the fancy looks I am after.

The biggest issue I have currently is that I have no idea how TLS works, so I don't know if my TOFU implementation is secure. I use rustls to deal with most of the TLS mess, but I needed to write a custom certificate verifier because it doesn't support the "trust on first use" system Gemini uses.

On the verifier, I am currently ignoring TLS v1.2/v1.3 signature verification, which I have no idea how to do manually, and wasn't able to find any sane documentation about it. The few pieces I have found suggest it requires the use of certificate authorities, which aren't relevant to TOFU, so I don't exactly know how to proceed, even if it was false and they were relevant I have no idea how to implement them.

Right now I am just making it succeed unconditionally. rustls has a default implementation for them I could use, but it breaks connection with some servers (particularly mine, something about certificate versions and subject alternative names I think), and that's why I am in this mess right now.

Any TLS wizards, the relevant code is here and any help is appreciated.

I ended up doing a bunch more work on my various map projections.

For the polyhedral projections, I implemented the 5 platonic solids, and made it so you can progressively subdivide them to more closely approximate a sphere. I also added code to allow you to rotate the map around without moving the faces so you can make the continents line up more optimally with the faces.

Then I went and implemented a bunch of azimuthal projections. I started with the "Little Planet" projection because I've seen it used a lot in artwork. But I found some really interesting ones like Azimuthal Equidistant and Lambert Equal Area. Both of these project the entire globe (minus the point on the opposite side) onto a circle, so you can see all the continents at once. I also implemented the quirky Peirce Quincuncial projection. I do not understand the math at all, but I found an implementation on the web and adapted it for fun. I just love the idea of literally fitting a round peg into a square hole mathematically.

I've been working on my map editor. https://github.com/Apostolique/Apos.Editor It's now at a really nice spot. To recap the main features:

• snip I can select stuff behind other stuff.
• snip I can select stuff, deselect stuff or add stuff to existing selections.
• snip I can reorder stuff.
• snip I can copy paste.
• I can undo and redo.
• snip I can resize relative to a smaller subset of an object.
• I can snap to the grid.
• snip I have an adaptive grid when I zoom out. The grid can also be locked if needed.

Now that I have all that in place, I started finalizing my UI library. https://github.com/Apostolique/Apos.Gui This is what I'll be working on for the rest of the week. I'll have to write some docs and hopefully next week I can release a new stable version.

Here is an example UI snip. The code for that looks like:

Panel.Push().XY = new Vector2(100, 100);
if (Button.Put("Click me").Clicked) _toggle = !_toggle;
if (_toggle) {
    Panel.Pop();
    Panel.Push().XY = new Vector2(300, 100);
}
Label.Put($"Your name is '{_name}'");
Panel.Pop();

This supports reparenting UI components. That means a components gets teleported to a different parent.

I've been streaming almost every day to https://glimesh.tv/Apos. I'm liking the site so far. It's using the FTL protocol which is insanely fast. If I watch my own stream and I talk there is barely any echo. It's still small but they are working really fast to fix all the bugs. I had an issue the other day, I got a response within 20 seconds and it was fixed within 40 seconds. (My stream was shown as live even though I was done streaming.) Then a full site fix was done before the next day.

I worked on many other projects, but those are the ones I'm most proud of.

chromecast_mpris is a daemon for controlling Chromecasts from Linux that I've added a few new features to, including a new dark mode icon and support for displaying more metadata. Having trouble replicating this issue, though. Also, I found out that Linux Uprising took the time to write about chromecast_mpris, which is pretty cool.

I added an async API to play_sounds and better support for making sure that spawned player processes are killed when the parent process doesn't shut down cleanly. play_sounds provides a simple API over the cross-platform idiosyncrasies that Python's sound playing ecosystem presents.

I'm writing documentation and fixing bugs on aiopath, an async pathlib replacement for Python.