2025 September Community Call #2

[expede]

📆 Event Calendar Page

🍿 Video

YouTube

🪵 Chat Log

00:03:51	Elliot Evans:	I was there, didn’t realize diffie was there :O
00:04:03	Ivan Reese:	Oh hai Elliot!
00:04:15	Elliot Evans:	hiya
00:07:01	Peter van Hardenberg:	I am in denial about the end of summer
00:07:51	Elliot Evans:	Reacted to "I am in denial about..." with 👍
00:08:12	Vincent Van Dijck:	Reacted to "I am in denial about..." with 🍂
00:08:58	Peter van Hardenberg:	Alex with the fancy slideware
00:11:45	Brooke Zelenka (@expede):	Reacted to "I am in denial about..." with 🍂
00:13:13	Peter van Hardenberg:	Feedback on that API is welcome, particularly from anyone who has rolled their own solution.
00:15:27	Peter van Hardenberg:	(I’m personally interested in exploring whether a system like this could be used for moderation, but that’s a someday project.)
00:16:03	megan w:	Reacted to "(I’m personally inte..." with ❤️
00:16:39	Rob Elves:	Replying to "Feedback on that API..."

I’ve been using the announce policy in samod which is working well…not sure if its the ‘new way’ or…
00:17:00	Shu Cao:	Just checking my understanding: if revocation happens concurrently with changes made by Alice and Bob on top of Alice's, does that mean both the new Alice and Bob changes won't be visible anymore?
00:17:14	Shu Cao:	Sorry this is my first intro to Automerge, so prob stupid!
00:18:56	Elliot Evans:	Do my changes depend on someone else’s if I just type in some text after seeing their changes?
00:19:03	Alex Good:	Replying to "Do my changes depend..."

Yes
00:19:09	Elliot Evans:	Reacted to "Yes" with 👍
00:19:32	Peter van Hardenberg:	Feel free to ask more questions here in the chat.
00:19:42	Peter van Hardenberg:	@Alex Good do we have a plan for viewing the ILLEGAL changes?
00:19:51	Peter van Hardenberg:	.view(WITH_CRIMINAL_INTENT)
00:20:12	Elliot Evans:	Replying to "Do my changes depend..."

If some changes are revoked after I saw them, I assume any future changes I add will not depend on those revoked changes?
00:20:17	Peter van Hardenberg:	Replying to "Do my changes depend..."

correct
00:20:24	Alex Good:	Not yet, but we probably should
00:20:29	Peter van Hardenberg:	Replying to "Do my changes depend..."

Changes are always dependent on the state of the document when you make them
00:20:34	Elliot Evans:	Replying to "Do my changes depend..."

So revocation affects a bit more than just the materialized view
00:20:35	Alex Good:	Reacted to ".view(WITH_CRIMINAL_..." with 😀
00:20:36	Elliot Evans:	Replying to "Do my changes depend..."

Ah I see
00:20:39	Elliot Evans:	Reacted to "Changes are always d..." with 👍
00:20:45	Elliot Evans:	Reacted to "Changes are always d..." with 👀
00:20:50	Peter van Hardenberg:	Replying to "Do my changes depend..."

(You can always backdate your changes with changeAt())
00:20:54	Elliot Evans:	Reacted to "(You can always back..." with 👍
00:21:00	Elliot Evans:	Reacted to "(You can always back..." with ❤️
00:21:21	Elliot Evans:	Reacted to ".view(WITH_CRIMINAL_..." with 😀
00:21:24	Peter van Hardenberg:	Replying to "Do my changes depend..."

(Which notably is not a way of escaping revocation because those changes would be concurrent with the future.)
00:21:48	Alex Good:	Replying to "Do my changes depend..."

Orion thinks that we might  be able to treat unauthorized changes as deletions (conceptually, not in the code) and thus downstream changes which are not rendered invisible by deleting the ancestor changes (e.g. editing siblings in a map) might continue to be visible
00:22:07	Elliot Evans:	Would be interested in exploration of UX on how to surface to users that changes they saw were revoked
00:22:26	Peter van Hardenberg:	Replying to "Do my changes depend..."

In some specific cases this would work! In others… not so much. Maybe we can consider this a kind of annotation in the new diff APIs… hmmm.
00:22:30	Elliot Evans:	Reacted to "Orion thinks that we..." with 👀
00:22:36	Elliot Evans:	Reacted to "In some specific cas..." with 👀
00:22:36	Adrien:	Reacted to ".view(WITH_CRIMINAL_..." with 😀
00:23:17	Peter van Hardenberg:	Replying to "Would be interested ..."

That would be great. Paul is doing some new work on general purpose annotation APIs for Automerge that might be a good building block. Maybe that’ll be ready for next month’s call but you can also just DM me and we can share
00:23:30	Elliot Evans:	Reacted to "That would be great...." with ❤️
00:24:42	Peter van Hardenberg:	I will note that those 200k documents are hosted on a small digital ocean droplet
00:25:12	Rob Elves:	Reacted to "I will note that tho..." with 😁
00:25:30	Peter van Hardenberg:	Using nothing more than the node-fs storage plugin
00:25:52	Peter van Hardenberg:	Though some Italian newspaper is pushing 30,000,000 change documents on a semi-regular basis that are making the sync server sad
00:26:17	bigmistqke:	Reacted to "Though some Italian ..." with 😂
00:26:36	Rob Elves:	Reacted to "Though some Italian ..." with 😂
00:27:34	Yuriko Nishijima:	this is probably a very very basic question about a causal graph, but when would it have a diverged change graph (having multiple child from one parent) in one device/peer?
00:27:52	Alex Good:	Replying to "this is probably a v..."

E.g two CLI applications using the same filesystem storage to modify the same document
00:27:59	Peter van Hardenberg:	Replying to "this is probably a v..."

You could also use changeAt() to backdate changes, Yuriko
00:28:38	Peter van Hardenberg:	“Old data is cold data."
00:28:41	Carlos Scheidegger:	love that hash trick, a genuinely useful thing I first learned from through the bitcoin paper!
00:29:41	bigmistqke:	Sort of an LOD
00:30:44	Alex Good:	Replying to "Sort of an LOD"

Not a perfect metaphor because each chunk does contain all the data - it's not a lossy compression. It's a bit more like levels in an LSM-tree
00:30:51	bigmistqke:	Reacted to "Not a perfect metaph..." with 👍
00:31:48	Peter van Hardenberg:	HTTP is nice because you can run it in the service worker’s background sync protocol which (last I checked) does not have websocket support
00:32:10	Elliot Evans:	n peers in n roundtrips? Like total?
00:32:28	bigmistqke:	I almost got its first GitHub star, but someone beat me to it
00:32:39	chee:	Reacted to "I almost got its fir..." with ❤️
00:32:39	walkah:	Reacted to "I almost got its fir..." with 👀
00:32:41	megan w:	Reacted to "I almost got its fir..." with ❤️
00:32:45	Aris Goudouras:	Reacted to "Though some Italian ..." with 😂
00:32:50	walkah:	Replying to "I almost got its fir..."

First!
00:32:57	Brooke Zelenka (@expede):	Reacted to "First!" with ⭐
00:33:12	bigmistqke:	Replying to "I almost got its fir..."

🏆 congrats!
00:33:16	bigmistqke:	Reacted to "First!" with ⭐
00:33:17	walkah:	Reacted to "🏆 congrats!" with 🙏
00:33:27	chee:	Reacted to "First!" with ⭐
00:34:12	Elliot Evans:	Yea, that seems… pretty dang good?
00:34:24	Peter van Hardenberg:	Reacted to "Yea, that seems… pre..." with 🎉
00:34:25	bigmistqke:	Reacted to "Yea, that seems… pre..." with 🎉
00:34:26	Brooke Zelenka (@expede):	Reacted to "Yea, that seems… pre..." with ❤️
00:34:27	walkah:	Reacted to "Yea, that seems… pre..." with ☝️
00:34:31	walkah:	Reacted to "Yea, that seems… pre..." with 🎉
00:35:01	Elliot Evans:	Reacted to "love that hash trick..." with 💰
00:36:12	Brooke Zelenka (@expede):	Contact cards are pretty ugly right now just to get it working. We plan to make them more palatable (QR codes, etc)
00:36:23	bigmistqke:	Reacted to "Contact cards are pr..." with 👍
00:36:31	walkah:	Reacted to "Contact cards are pr..." with 👍
00:36:39	Peter van Hardenberg:	PULL is for your sync server / a less-trusted message carrier
00:36:59	Brooke Zelenka (@expede):	Reacted to "PULL is for your syn..." with 👆
00:37:46	bigmistqke:	I think u gotta copy contact card again
00:37:53	bigmistqke:	U missed a }
00:39:11	Peter van Hardenberg:	Observe that you can also choose per-document which servers you sync to!
00:39:17	Brooke Zelenka (@expede):	Reacted to "Observe that you can..." with 💯
00:39:23	walkah:	Reacted to "Observe that you can..." with 🔥
00:39:34	Rob Elves:	Reacted to "Observe that you can..." with 💯
00:39:41	Dima:	Does access work transitively with groups, i.e. can I create a group for my team, give all docs admin access to that team, and then manage members of that team
00:40:05	David Luecke:	What was in the “contact card” again?
00:40:05	Dima:	(So that I don’t have to change permissions for every document I share with my team)
00:40:30	Elliot Evans:	Transitively both by groups and by folders?
00:40:35	Peter van Hardenberg:	Replying to "Transitively both by..."

Yes
00:40:57	walkah:	Reacted to "What was in the “con..." with ☝️
00:41:03	Adrien:	Can roles be arbitrarily defined? Not limited to admin/write/read/pull?
00:41:04	Elliot Evans:	Reacted to "Yes" with 🔥
00:41:56	Dima:	Can I reuse keyhive structure to give application-defined roles to group members?
00:42:01	Adrien:	Replying to "Can roles be arbitra..."

Indeed, I'm quite new here so I don't have much context haha.
00:42:06	Elliot Evans:	Replying to "Can roles be arbitra..."

you can get more granular in a single document by breaking the larger document Ito multiple smaller documents
00:42:48	walkah:	Replying to "What was in the “con..."

Bump this one. Are contact cards required? Documented?
00:43:18	Adrien:	Replying to "Can roles be arbitra..."

Ok I see! Thanks!
00:43:20	Brooke Zelenka (@expede):	Replying to "What was in the “con..."

They’re just the Ed25519 public key + X25519 encryption agreement key
00:43:27	David Luecke:	Replying to "What was in the “con..."

I’m assuming it’s at least a public key but it looked like it had some other info.
00:43:36	bigmistqke:	It’s cool that the sync server is also a user in the system with its own role
00:43:44	David Luecke:	Replying to "What was in the “con..."

Ah ok that makes sense
00:43:47	walkah:	Reacted to "It’s cool that the s..." with 💯
00:43:54	Dima:	Got it. I will think about concrete use-cases and reach out to best model them with this design. Thanks!
00:43:56	chee:	Reacted to "It’s cool that the s..." with 💯
00:43:57	walkah:	Reacted to "They’re just the Ed2..." with 🙏
00:44:06	Elliot Evans:	Reacted to "It’s cool that the s..." with 💯
00:44:18	walkah:	Replying to "What was in the “con..."

I found https://github.com/inkandswitch/keyhive/blob/cdeba49fdccd3e6b1dfec0532f3f2a67ac84fab7/keyhive_core/src/contact_card.rs#L11
00:44:19	walkah:	Replying to "What was in the “con..."

🙂
00:45:14	Peter van Hardenberg:	Replying to "What was in the “con..."

I would treat the “contact card” as just a demo placeholder. There’s room within the design for lots of different approaches here: you could layer on a PAKE system or invite codes or various other mechanisms as well.
00:45:28	megan w:	Reacted to "It’s cool that the s..." with 💯
00:45:42	Elliot Evans:	Is access control granting/revocation always hierarchical, could you have two peers that can both revoke each other’s access?
00:46:09	bigmistqke:	Let’s go cheee
00:46:10	Peter van Hardenberg:	Replying to "Is access control gr..."

You can indeed create what’s known as a "revocation cycle”
00:46:22	Elliot Evans:	Reacted to "You can indeed creat..." with 😎
00:46:48	Peter van Hardenberg:	Replying to "Is access control gr..."

Figuring out a principled way of solving was a pernicious part of the design and is one of the big conceptual contributions from Brooke
00:46:57	Elliot Evans:	Reacted to "Figuring out a princ..." with ❤️
00:46:58	Brooke Zelenka (@expede):	Reacted to "Figuring out a princ..." with ❤️
00:46:59	David Luecke:	Reacted to "It’s cool that the s..." with 💯
00:47:07	Elliot Evans:	Replying to "Is access control gr..."

Lets go Brooke
00:47:12	Peter van Hardenberg:	(Or just ask things in the chat)
00:47:13	Brooke Zelenka (@expede):	Reacted to "Lets go Brooke" with 🚀
00:47:15	bigmistqke:	Reacted to "Figuring out a princ..." with ❤️
00:47:15	Ivan Reese:	Reacted to "Let’s go cheee" with 👏
00:47:17	Aris Goudouras:	Reacted to "Figuring out a princ..." with ❤️
00:47:52	Peter van Hardenberg:	(But why would you want that)
00:48:20	Peter van Hardenberg:	I’ll share it
00:48:41	Brooke Zelenka (@expede):	Replying to "Is access control gr..."

There’s a bit about this in the lab notes, but I’ve been meaning to also write up in more detail https://inkandswitch.com/keyhive
00:48:50	Ivan Reese:	Yes I want this also
00:48:53	Peter van Hardenberg:	https://github.com/automerge/automerge-repo/tree/main/packages/automerge-repo-solid-primitives
00:49:16	bigmistqke:	lol
00:49:19	Ivan Reese:	The video game Journey.
00:49:28	Elliot Evans:	Reacted to "The video game Journ..." with ❤️
00:49:29	Peter van Hardenberg:	Elden Ring?
00:49:45	Ivan Reese:	Replying to "Elden Ring?"

Nobody ever griefs in a From game.
00:50:18	Carlos Scheidegger:	Replying to "Elden Ring?"

i must be remembering very differemt messages dropped by other Elden Ring players than y'all
00:50:34	Brooke Zelenka (@expede):	Reacted to "i must be rememberin..." with 😛
00:52:24	bigmistqke:	🥳
00:52:25	chee:	https://slaygrounds.netlify.app/#automerge:3Bv5VtowYvU869J79oCfwoCy9V2y/entry.tsx
00:52:39	megan w:	Reacted to "https://slaygrounds...." with ❤️
00:52:44	Brooke Zelenka (@expede):	Reacted to "https://slaygrounds...." with ❤️
00:53:09	Ivan Reese:	Indeed, Automerge has solid support. And Solid support.
00:53:17	Alex Good:	Reacted to "https://slaygrounds...." with ❤️
00:53:18	bigmistqke:	Nice demo Chee!
00:53:20	Josiah:	Reacted to "https://slaygrounds...." with ❤️
00:53:21	bigmistqke:	Reacted to "https://slaygrounds...." with ❤️
00:54:12	Adrien:	Reacted to "https://slaygrounds...." with ❤️
00:54:24	Elliot Evans:	Reacted to "Nice demo Chee!" with 👍
00:55:08	Elliot Evans:	Reacted to "https://slaygrounds...." with ❤️
00:55:14	Shu Cao:	Reacted to "https://slayground..." with ❤️
00:55:25	Chris Boscolo:	This was great stuff! Good work everyone!
00:55:43	Chris Boscolo:	gotta drop
00:55:46	megan w:	really cool :))
00:56:06	chee:	Reacted to "really cool :))" with ❤️
00:56:08	bigmistqke:	👏👏👏 great work everyone!
00:56:23	Rob Elves:	👏
00:56:38	Shu Cao:	Thanks! That's really great!
00:56:44	chee:	can’t believe nobody else wore their Automerge-themed outfit, i thought this was a party
00:56:54	bigmistqke:	Reacted to "can’t believe nobody..." with 😂
00:56:57	Elliot Evans:	Reacted to "can’t believe nobody..." with 💛
00:57:08	Rob Elves:	Reacted to "can’t believe nobody..." with 😂
00:57:11	megan w:	Reacted to "can’t believe nobody..." with 💛
00:57:12	Brooke Zelenka (@expede):	Reacted to "can’t believe nobody..." with 💛
00:57:14	bigmistqke:	👋 see ya everyone
00:57:32	Adrien:	Thanks for the community call !!
00:57:38	Adrien:	See you next time!
00:57:50	Elliot Evans:	Thank you!!
00:57:53	Elliot Evans:	Great to see yall
00:57:54	nicabar nimble:	Thank You
00:57:55	Peter van Hardenberg:	Ciao all!
00:57:57	megan w:	thanks so much!!
00:57:57	Rob Elves:	Thank you!
00:57:58	John Mumm:	Thanks!

[joshmarinacci]

I'm bummed that I missed seeing this live. I'd like to hear more about Alan Kay showing up in previous meetings. He's a fascinating person to talk to.