October 2026 Meeting

The technical steering committee met on October 1. Ryan, Boris, Ms Boba, and Nick made quorum. We also had a bunch of first-time guests, which is a good sign that the meeting reminders and forum posts are reaching people.

Thanks to everyone that joined us live on call. It felt like when we first got started with tech talks and had some hang out vibes. People are welcome to join, introduce themselves, and talk a little bit about what they’re doing. If there’s interest we could plan ahead for hosting presentations at another time, our monthly TSC meetings only have so much room.

What follows is a prose style transcription / summary of the call that has been lightly edited for clarity or because we want to make a dedicated blog post about parts of it.

Nick also opened with the usual reminder: this is a public meeting, everyone in it is human, and people are allowed to feel strongly about lexicons without being a jerk about it.

Feedback is part of the job

Jedi Wright, new to both the group and AT Protocol, brought questions about using lexicons for local-first work, coming at it from a data sovereignty angle. The questions were already on the forum, and the ask was simple: more eyes. Boris is sharing the post with people he knows in the local-first world, and Nick put time on his calendar that evening to read it.

Ryan made a point worth repeating. Giving feedback on lexicons we'll never own is as much our job as shepherding the ones we do, and so is connecting people who are building overlapping things.

Text is hard, images are close

Alan Smith walked through a section-based markup format he's been building. Think markdown with attributes: content stays separate from layout, metadata can live in the text, and anything unformatted falls back to plain paragraphs. His case against plain text is that nobody agrees how to transform it. His case against markdown included spoilers, which force you to drop HTML into your content.

Ryan pointed him at the standard.site folks, who left the content format unspecified and have lived with both sides of that choice. Ms Boba reframed the question. In her view, the hard problem is how a lexicon carries more than one text representation, and picking a format comes after that. She asked Alan to start a forum thread from that angle. Nick wants the same thing from a different direction: a format-tagged string that can sit in an open union, so the record says what the text is and stays out of the format fight.

Boris noted that the book working group is still informal, a space on the forum more than a chartered group. If its co-leads want to bring their work into the namespace, they'll need a TSC sponsor.

The standalone image lexicon is in better shape. Discussion has mostly converged, naming is the main open question, and alt text (including multi-language alt text) is in. Ms Boba "volunteered" to sponsor it, and Boris offered thoughts and prayers. She also asked whether something this small needs a formal working group at all, and suggested we write down some guidance on scoping working groups. Boris's advice was to find implementers first. A shared image record is a big unlock, but only once apps actually use it.

Specs people are already shipping

Service self-description has moved from "how should we do this" to "we're doing this, come join." Blacksky, Happy View, atproto-pds, and a few others already implement it. The hard part now is getting Bluesky to ship something in their PDS that didn't start with them. The forum thread is open if you have opinions.

Nick also gave a quick tour of why CIDs make file sharing tricky. A CID tells you exactly which bytes you have and nothing about what they are. That leaves real questions about discovery, replication and streaming, and checking authorization in permissioned spaces. The DASL specs could use developer-experience feedback, so go read them and say something.

Who decides

Cheese Community has a co-lead now and is moving slowly on purpose. It may sound silly, but Boris is using it as a small, real test case: at some point he'll ask the TSC to oversee it, and we'll have to work out how a domain gets onboarded. That's the question the domain stewardship and namespace policy item keeps circling, so Cheese Community gets to answer it first.

The open social community groups work is still settling names and identifiers after a meeting the day before. Boris's questions were about governance. It's a lexicon the community will adopt widely and a core protocol primitive at the same time, so who are the co-leads, and how do they make decisions before anything heads to the IETF? He pointed to the recent DASL governance update as a lightweight model worth reading.

Ms Boba noticed Bluesky hasn't been represented at the last few meetings. Bryan had a genuine conflict this time, but the point stands, and several people at Bluesky have said they'd like to take part. Boris put the bigger goal plainly: multi-stakeholder arrangements, which means keeping an eye on large actors doing whatever they want, with no assumption of bad intent.

TSC membership and governance went to the backlog for November, with rotating seats as the main topic. Before then, Nick will check in with each TSC member about interest and how much time they can give. Time zones are part of it, since the meeting lands in Tom's already busy evening. Nick floated making one meeting a year (maybe November) mandatory for TSC members, and Ryan suggested alternating between a slot that works for Asia and the US West Coast and one that works for Europe.

Any other business

A few guests introduced themselves. Jarret Redding comes from podcasting, got tired of waiting on Podcasting 2.0, and thinks AT Protocol can give podcasts the social layer they've never had. Alan, Al9000 on the forums, is about two months into AT Protocol. Florian has a podcast app prototype too, and he's back on atmo.rsvp and the community events lexicon, where by his count about half the fields in use aren't in the spec. Boris admitted that he, Guido, and Ms Boba are about to invent a pile of event fields for AtmosphereConf, so the better plan is to sort those out together ahead of time. Ms Boba asked Florian to post examples of his custom fields to the forum to get that going.

Nick is planning a small lexicon unconf before AtmosphereConf. It's a meet-and-greet for anyone who makes lexicons, uses ours, or thinks something belongs in the namespace, not a how-to-contribute session. Boris gave it a tentative yes, and details will follow as the schedule firms up.

Nick also showed off Highport (highport.space), which hosts sites defined as lexicon records. A PDS can delegate a wildcard domain to it, give every handle a default site, and register handles through an assignHandle call during account creation, before the PLC operation goes out. A lexicon for that method is coming. Ryan, who has spent a couple of decades around the IndieWeb, said the big usability jump there came from registrar and DNS host APIs, and existing APIs are a better bootstrap than a new standard nobody has adopted yet. Nick agreed, with the caveat that every DNS provider's API is different and all of them are a pain.

The last few minutes drifted into notifications. The AtmosphereConf newsletter went from about zero to somewhere between 300 and 400 subscribers, roughly 80 of them by email, and Boris isn't confident the rest will see updates in time through standard.site readers. For most people, notifications means their phone. Boris wants to use AtmosphereConf's live audience as a testing ground for notification implementations, and Ms Boba is prototyping notifications on her own site that start from the user and their PDS instead of the app. Nobody has this one figured out. In the meantime, Boris is writing a post on tuning your Discourse notifications for everyone drowning in the forum.

Next meeting

The November agenda already has one item: TSC membership, including rotating seats and what we expect from TSC members.

November 5, Thursday, 1 PM ET / 18:00 UTC. Forum. Show up if you want something discussed.