This FEP introduces a standardized way for instances to advertise stores of “governable events,” or activities for moderators to take action on, and for clients query them to power BI visualizations and moderator investigations.
This system takes a lot of inspiration from ROOST’s Osprey system for moderation rules and investigations, and this FEP aims to fill the Druid-sized hole in powering that kind of dashboard on the Fediverse.
I think it’s critical that we start thinking about how to offer robust moderation systems standardized for C2S so clients can implement moderation UIs instead of forwarding that to a non-standard UI provided by the instance. This is something critical to every instance, and standardizing it to create a real trust and safety ecosystem and culture makes sense to me.
Please read the FEP, then provide feedback at https://codeberg.org/fediverse/fep/src/branch/main/fep/c0a8/fep-c0a8.md !
Aside: looks like the FEP topic is read-only so I submitted it here.
One thing stands out to me: The objective of ActivityPub open standard is to enable autonomous parties in a technology ecosystem to develop interoperable social networking apps and services.
In that regard:
- “Governable events” → Modeling of potentially useful interoperable concepts in a “Moderation” or “Governance” busines domain.
- “Governable events” implementation → Prescribing a set of additional specifications requiring a whole different tech-stack to implement these concepts. These specs go an entire different direction. It’s like adding a different world to AP.
It’s a small wrapper over existing Activity Streams to add context around automated policy decisions, I don’t see how it’s a different direction. Do you have any suggestions around how to better integrate it with the existing AP world?
The wrapper is that governableEvents is the single property entry point to a whole business domain that will evolve in a very different techstack. You model the domain in different technology entirely, so anyone who wants to interoperate on your moderation concepts needs to follow suit. The FEP is stacking the following at that endpoint as a requirement to interact with the business domain:
Apache Arrow is a universal columnar format and multi-language toolbox for fast data interchange and in-memory analytics.
Arrow Flight is an RPC framework for high-performance data services based on Arrow data, and is built on top of gRPC and the IPC format.
Arrow Flight SQL is a protocol for interacting with SQL databases using the Arrow in-memory format and the Flight RPC framework.
gRPC is a high performance, open source universal RPC framework.
Substrait is a format for describing compute operations on structured data. It is designed for interoperability across different languages and systems.
You may of course go ahead with this FEP, but it is not very attractive I’d imagine for other ecosystem parties to implement it. People who stick closer to the AP techstack will model Moderation / Governance their way.
Suggesting possible ways to improve the FEP ..
-
Define a bridge. Make the endpoint descriptive for what it offers i.e. an apacheArrowRPC endpoint. That is straightforward and clear, albeit more generic and broader scoped.
-
Elaborate more of the “Governable events” domain using AS/AP with JSON-LD linked data msg formats and business logic, and then describe a non-normative mapping to Apache Arrow et al.
Gotcha, my impetus for this usage of Arrow/Flight was to integrate with existing BI tooling, ADBC clients and OLAP databases. For instance both ClickHouse and DuckDB, two popular and extensible OLAP stores, both support Arrow and/or Flight SQL, making it trivial to expose them in a standardized fashion to moderators for querying under this specification.
I’m worried that modeling it closer to the existing ActivityPub tech stack would break too much compatibility with the domain being targeted, that of investigation tools and other BI software.
Your points though about making the fact that it’s using Flight more transparent and modeling governable events with JSON-LD are well taken though, and I’ll update the FEP.
I am not too familiar with Arrow/Flight and BI tooling, but from what I’ve seen there are thriving technology ecosystems around the specifications, so I think I think it could be very valuable to have a FEP for the general case and define a bridge. If you choose that route I think the FEP title and thus ID would have to change, I guess.
I do think though that it would be valuable to standardize around that existing domain rather than it just being a ‘bridge’. There is no AP model for governance, and I think building out an AP-native model for it would end up being a hinderance for adoption rather than a support.
For instance, Bluesky/ATProto took the route of prescription when supporting moderation databases by explicitly recommending the use of Osprey which requires the use of Druid for event storage and uses an idiosyncratic API: The AT Stack - AT Protocol
This has led to the stalling of development around an ecosystem beyond that of Osprey for Bluesky moderation, with no alternative clients, OLAP stores, etc. I think that kind of approach would be a mistake and we should be working within the already thriving BI domain when designing our moderation pipelines to build a thriving ecosystem.
There has been a lot of discussion on the subject. I am not sure about what work has been done already in this regard, or what people have in the planning. Not sure about overlaps with SWICG Trust & Safety and IFTAS, for instance (cc @thisismissem).
However, I would say that - given how pervasive in scope the FEP currently is, in what it prescribes - I think a FEP wouldn’t be the best approach to get broad support and adoption, and I’d go the route of W3C SWICG. Might start discussing the matter on the SWICG mailing list.
You can’t really compare to ATProto in this regard, because it has a way more centralized team of paid employees that drive the specs, and immediately take care that testsuites and impls follow suit. AS/AP is much more grassroots, which makes getting buy-in more involved and difficult.