-
[email protected] yes, Mastodon compatibility is lacklustre but that's partly because the architecture of Mastodon doesn't lend itself well to categorical organization. That's not a criticism, merely a difference that we have to consider.
Currently, when a user addresses multiple communities, then the existing software (NodeBB included) uses the first one. How the rest of the addressed communities are handled is what's of interest here.
@julian @kirkmoodey Mastodon is currently working on a spec for their groups implementation. Maybe get everybody together to hash out a common spec, including those who already have a group implementation like Friendica/Hubzilla, Misskey/IceShrimp/Sharkey, Pleroma/Akkoma?
-
IMO The simplest way would be to garnish a bit of extra data onto the normal FEP 1b12 process.
Create a new post (Create -> Page to the instance that hosts the community, which in turn does Announce -> Create -> Page to followers) and add an extra field to the
Pagewhich is the URL of the original post. That will establish the association.To reject the cross-post, return HTTP 400 (403?) to the POST to the inbox on the initial Create -> Page ? Or send a Reject activity, either way is fine but the 400 seems easiest. Lemmy returns 400 for a lot of things so we have some prior art in that direction.
-
@julian @kirkmoodey Mastodon is currently working on a spec for their groups implementation. Maybe get everybody together to hash out a common spec, including those who already have a group implementation like Friendica/Hubzilla, Misskey/IceShrimp/Sharkey, Pleroma/Akkoma?
gunchleoc:Mastodon is currently working on a spec for their groups implementation.
Any links to this and/or discussion of how it relates to other FEPs?
-
[email protected] yes, you're right about that. The silver lining here is that we're at a point where the major players' implementations are local-only, so we don't have to work around with a pre-existing implementation and ensure compatibility.
The second point is that a "cross-post" could have multiple meanings, including manually creating a new post about a link already present in multiple communities. What I'm hoping to describe is a common way that items can be cross-posted natively between instances, while hopefully preserving their reply-trees.
Duplicating the object would mean the discussion is split between objects. The ideal implementation would be the same object present in multiple categories/communities. Is there desire for this in the threadiverse?
If the link goes to a controversial news article and it's get posted into pro- and against- community/group the comments will spiral out of control and it won't be a pleasant place.
Maybe it could be implemented as a toggle per group/instance within one fedi software. It shouldn't be in Activity Pub protocol.
-
gunchleoc:
Mastodon is currently working on a spec for their groups implementation.
Any links to this and/or discussion of how it relates to other FEPs?
-
THanks ... turns out I knew about that: the implementation for the NLNet grant, but never released. My impression is that it's been on hold since then, and there's so much other discussions of group-releated FEPs that I certainly hope they'll incorporate newer thinking if and when it moves forward.
-
THanks ... turns out I knew about that: the implementation for the NLNet grant, but never released. My impression is that it's been on hold since then, and there's so much other discussions of group-releated FEPs that I certainly hope they'll incorporate newer thinking if and when it moves forward.
I haven't looked into the differences between their implementation and how groups are implemented using 1b12, but what I have discovered is that the 1b12 community is much larger than I gave it credit for.
-
P projectmoon moved this topic from ActivityPub
-
How would this work on the NodeBB side? Multiple categories associated with one topic?
-
How would this work on the NodeBB side? Multiple categories associated with one topic?
[email protected] basically, yes. It would be a little too involved to upend the entire system to support multiple cids per topic — a lot of our existing code relies on
cidbeing a single value.This would be an add-on logic of sorts, where each topic has a canonical category, but can also be cross-posted to other communities/categories.
-
IMO The simplest way would be to garnish a bit of extra data onto the normal FEP 1b12 process.
Create a new post (Create -> Page to the instance that hosts the community, which in turn does Announce -> Create -> Page to followers) and add an extra field to the
Pagewhich is the URL of the original post. That will establish the association.To reject the cross-post, return HTTP 400 (403?) to the POST to the inbox on the initial Create -> Page ? Or send a Reject activity, either way is fine but the 400 seems easiest. Lemmy returns 400 for a lot of things so we have some prior art in that direction.
Hey [email protected] thanks for responding (and sorry for the late reply!)
I am not married to the
Announce([Article|Note|Page])approach, so I am definitely open toCreate([Article|Note|Page])with a back-reference. I think I went the former direction because there is a known fallback mechanism — theAnnounceis treated as a share/boost/repost as normal. However, sending theCreatealso is fine I think.- However, do we need a backreference? In my limited research, it seems that Piefed, et al. picks the first
Groupactor and associates the post with that community. If I sent over aCreate(Article)with twoGroupactors addressed, could Piefed associate the post with the first, and initiate a cross-post with the remainingGroupactors? - Secondly, is how to handle sync. 1b12 relies on communities having reciprocal followers in order for two-way synchronization to be established. On my end since I know it is cross-posted I will now send
1b12activities to cross-posted communities, but can Piefed, et al. send 1b12 activities back as well, in the absence of followers?
cc [email protected] [email protected] [email protected] [email protected]
- However, do we need a backreference? In my limited research, it seems that Piefed, et al. picks the first
