With an FPGA or special CPU instruction set, the encryption algorithms could run on a toaster—which would give access to whatever low-spec handheld you wanted without making it chug to have strong encryption. That also still isn’t covering the future hope of a Linux phone, or someone that just wants to register an account on their laptop.

Using forks puts stress on other teams to keep up with breaking changes, & 90%+ of folks won’t be looking for forks or be willing to trust their unofficial status. I saw the code for UnifiedPush as a Mattermost plugin & it was like 50 lines or something small which is much less than the rest while allowing users to keep control of their metadata which is a big deal if you care about privacy. A fork for SMS support would encounter similar issues, & now you either need to compete with Molly or copy its featureset otherwise users have to choose, SMS or UnifiedPush. That said, I agree with the SMS situation since it was easy to convince relatives to use this new “text app” where encryption magically came to a chunk of their contact list.

Saying emoji was the most important was tongue-in-cheek, but it makes the application feel non-native (& I think Apple’s emoji are particularly ugly). You would think at least the Google set was shipped to Android, or—now hear me out—not ship emoji, don’t override the user experience, let the user’s fontconfig display the one they set. Shipping a whole font (or images) for emoji is why the application size is so bloated for a chat app.