Video
Chat Log
00:05:03 Brooklyn Zelenka (@expede): Not one, but TWO PvHs!
00:05:10 Peter van Hardenberg: coming at you in stereo
00:07:11 Peter van Hardenberg: interactive text editing makes this problem even worse
00:07:16 Peter van Hardenberg: 300b per character
00:07:23 Sodium : ouch
00:14:41 Sodium : https://github.com/fission-codes/spec/pull/12 ?
00:16:23 Peter van Hardenberg: nice to see you in the call, zeeshan
00:16:32 Zeeshan Lakhani: Nice to see you again PVH 🙂
00:18:19 Sodium : for later, the demo from last month was recorded, and posted here: https://talk.fission.codes/t/fission-reactor-august-2022-research-update/3257
00:19:30 Sodium : https://github.com/vmware/database-stream-processor
00:19:49 Sodium : https://github.com/vmware/differential-datalog (i think)
00:19:58 Marc-Antoine Parent: Yes
00:20:05 Brooklyn Zelenka (@expede): Wow, you're really on these links today Sodium 🙌
00:20:23 Marc-Antoine Parent: Also https://github.com/TimelyDataflow/differential
00:20:31 Sodium : i just like it when sources are cited :P
00:20:38 Brooklyn Zelenka (@expede): We dread DRed
00:24:05 Peter van Hardenberg: we've got a new dataset coming -- martin's been using a plugin we wrote in vscode for a new paper
00:24:15 Brooklyn Zelenka (@expede): Nice
00:24:30 Peter van Hardenberg: well done, quinn!
00:24:38 Brooklyn Zelenka (@expede): 🥳
00:24:58 Sodium : :tada:
00:27:35 Quinn Wilton: Thanks Peter!
00:28:30 Randall: https://arxiv.org/abs/2203.16684
Is this the doi for the DBSP paper?
00:28:39 Quinn Wilton: That's the one!
00:28:43 Quinn Wilton: Their implementation is here: https://github.com/vmware/database-stream-processor/
00:29:11 Quinn Wilton: I've been super impressed with the system. It's essentially a simpler formalization of Differential Dataflow catered to relational algebra
00:29:20 Peter van Hardenberg: sorry -- some latency, apologies for the interruption
00:30:00 Peter van Hardenberg: this is where we wound up with the sync protocol with the bloom filters
00:30:12 Peter van Hardenberg: 1% chance of an extra roundtrip
00:30:45 Peter van Hardenberg: i wonder if you could adopt our opset/sync protocol
00:31:20 Brooklyn Zelenka (@expede): PvH — yeah that’s the one
00:31:33 Brooklyn Zelenka (@expede): Oh interesting, do you have a link to your sync protocol?
00:34:52 Peter van Hardenberg: alex has a spec
00:34:59 Peter van Hardenberg: he's been looking for an excuse to finish it
00:35:15 Sodium : probably particularly relevant is that wasm adds a bunch of restrictions/requirements around memory management, which is probably important for performant stuff
00:35:49 Peter van Hardenberg: https://github.com/automerge/automerge.github.io/blob/5b0d0284e78da92c95db6c2a3663e6990fd5d003/docs/how-it-works/sync.md
00:35:57 Brooklyn Zelenka (@expede): Thanks! 👀
00:36:00 Peter van Hardenberg: this is an informal description
00:36:56 Sodium : ah! yeah, also threading is a thing that can throw a wrench in wasm stuff
00:37:31 Zeeshan Lakhani: This is the paper on shared arragnements: https://people.csail.mit.edu/malte/pub/drafts/2019-kpg.pdf
00:37:39 Zeeshan Lakhani: Most def. Na.
00:37:52 Brooklyn Zelenka (@expede): PvH: Ah, right, this is with a Bloom filter acting as a vector clock! Ours is a bit more general for syncing arbitrary graphs on blockstores, partly because the remote store likely can’t read the links and the facts are shuffled
00:42:37 Brooklyn Zelenka (@expede): https://github.com/fission-codes/spec/blob/86567141cec9538077154fd5709f90464d8b6861/car-pool/car-mirror/SPEC.md
00:43:51 Quinn Wilton: I know that feeling 🙂
00:49:12 Quinn Wilton: https://ipld.io/specs/transport/graphsync/
00:51:04 Sodium : just have to make datalog and/or logic programming popular first
00:51:05 Sodium : easy
00:51:08 Sodium : :P
00:52:12 Zeeshan Lakhani: Na, lolz.
00:52:17 Zeeshan Lakhani: The “long” game.
00:53:44 Brooklyn Zelenka (@expede): haha
00:53:46 Sodium : what can i say? i like thinking about the social angle, even if that ends up being occasionally impractical
00:54:12 Quinn Wilton: There's a quote I like in a Datalog paper about how Datalog is to query languages what lambda calculus is to functional programming
00:54:42 Sodium : ooh, that's a nice way to think about it, and definitely fits
00:55:10 Quinn Wilton: It has such simple semantics with so many extensions that you can think of so many different query languages as being constructed from different combinations of datalog extensions, with a higher-level syntax
00:56:59 Randall: https://databasetheory.org/sites/default/files/2016-06/wood.pdf
The reference section is good for those wanting a survey of the GQL branch of the family …Tree 😁
00:57:14 Peter van Hardenberg: two weeks to two years, then
00:57:27 James Walker (@walkah): it'll be released by Christmas …
00:58:13 Peter van Hardenberg: TBD
01:00:50 Peter van Hardenberg: thanks for a great session quinn -- must head for airport shortly
01:00:53 Randall: Thanks 👏
01:00:57 Sodium : o/
01:00:58 Quinn Wilton: Thanks for joining! Have a nice flight
01:01:40 Zeeshan Lakhani: Lots of interesting implications details to always get right too. I’m thankful we have wasm people 🙂.
01:02:41 Jack Park: Great conversation. Thanks!
Register for the Event Series
Resources & Notes
Early spec description with high level wording is starting to go into the Fission specs repo: Dialog Specs by expede · Pull Request #11 · fission-codes/spec · GitHub