~

Roadmap

Update

~2023.2.19

Author

Ted Blackman

@P

~rovnys-ricfer

We did a small release on Tuesday, with no breaking changes. I'm happy that the release process only took three hours, down from ten hours a month ago. (~dinleb-rambep) is now streamlining it further.

I am pleased to announce that solid-state subscriptions are now ready for use. This is a userspace pub/sub system that should be significantly easier to use correctly than normal Gall subscriptions. The sss library is now in the "develop" branch of the urbit/urbit repo. ~wicrum-wicrun wrote thorough documentation for how to use the library. Before release, he also refactored its internal synchronization logic to use the "yore/nigh" scry response logic that ~wicdev-wisryt originally suggested for use in the Ad Fontes networking proposal.

On Friday we merged 9 PRs. It feels good to be building up speed in this way. One of those had been made less than a day before: support for the runtime injecting Hoon code to be run as a monadic script ("thread"), which will be useful for hosting providers and other Urbit orchestration systems. Community attendance at our daily triage meetings continues to be pleasant and helpful.

~dachus-tiprel and other community members have been looking into an issue that causes Urbit's HTTP clients to fail to reconnect. They've now narrowed it down to a client-side issue, likely due to improper error handling. It's good to be making progress on this bug, since it has impacts user experience directly. It's also nice to see the community work together to address the issue. As Linus Torvalds says, with enough eyes, all bugs are shallow.

~barter-simsum's pointer compression work, which increases Urbit's storage capacity to 8GB, is now working. We will review it in more detail this week, with plans to merge it the next week and deploy it one week after that.

There's been increasing activity on the Ares project lately. See, for instance, ~wicdev-wisryt's branch where he's been adding jets to the new interpreter.

Finally, I unearthed an old document describing the steps involved in the libames project. This would allow phones and embedded devices to speak Urbit's "Ames" peer-to-peer protocol, without needing to run a full Urbit node on the client. Since these clients currently communicate with their Urbit over HTTP, a user a user needs to take a bunch of rather technical actions to access their ship from their phone: buy a domain, set up a relay system using TailScale or NativePlanet's StarTram, and configure SSL termination. None of these activities sparks joy.