Community issue tracker

Community
1

This topic is for tracking technical issues related to SocialHub. Some of these issues will be related to the software itself and may lead to separate topics in Software > Discourse . Some may lead to other types of discussions. This topic is intended to be a living wiki to keep track of things, and also to share when I anticipate having time to address them.

Issues

Please feel free to amend this list.

  • Many topics in the FEP category are still not federated: Fediverse Enhancement Proposals - SocialHub .

  • Outbox is not paginated and returns a huge JSON document. In some cases it doesn’t work at all, returns 502, for example: https://socialhub.activitypub.rocks/ap/actor/271cb1ad85bf02c266edf53586d04254/outbox

  • Embedded Activity.actor makes Discourse incompatible with many Fediverse services.

  • An ability to create a post from a remote server by @mentioning a category.

  • Redirect from a post to a corresponding ActivityPub object when Accept header is set to application/ld+json; profile="https://www.w3.org/ns/activitystreams"

  • Federated threads end up in Community (based on hashtag?) but shouldn’t most likely. E.g. this topic.

  • General ActivityPub Space discussion moved to #Fediversity

Schedule

My next work slot for ActivityPub stuff is September 13-14.

5 Likes
2

Just a note, partially for myself, that I didn’t have the time I was expecting this past weekend (I ended up having to work on a job). I’m going to try to turn to this list again this coming weekend (September 20-21).

2 Likes
3

There has been a spam attack on the SocialHub, where about ~340 spammy topics were created and are flooding the recent activity feed. I managed to flag about 30 posts before the Discourse software rate-limited me and prevented me from flagging anything else for the next 24 hours.

We have two tasks:

  1. Clean up the existing spam

  2. Look into countermeasures we can take so this doesn’t happen again?

2 Likes
4

Much appreciated @trwnh :folded_hands:

I’ve stemmed the current tide. I will look at countermeasures asap.

1 Like
5

Just for transparency, this is what I did after @trwnh helpfully notified me about the spam issue.

  1. Temporarily required trust level 1 to post a new topic to prevent new posts being made by bots.
  2. Deleted the spam posts.
  3. Replaced the blanket tl1 requirement with a < tl1 post approval requirement (posts go into a “review queue”
  4. Set up AI-based spam detection using Pavilion’s Anthropic account, specifically using Claude 3.5 Haiku.
  5. Removed the blanket tl1 post approval requirement.
3 Likes
6

I believe the spam detection will fail due to the spammers specifically having a countermeasure for this – the spam posts are being made with innocuous text at first, then edited to include the spam.

7

Yeah, I think you’re right. I’m going to require the first post from tl0 (new) users to be approved for now. I need to go to various meetings soon and need to have something in place for the time being.

1 Like
8

Im a little concerned about AI for detecting all users.

Perhaps you can consider restricting it to trust level 1 users?

FWIW, my free time is going to be brutal until the Summer.

If you need a background moderator to help during random evenings (rather than procrastinating) feel free to add me.

1 Like
9

Yeah I hear you. I don’t love it either. It was more a question of potentially saving time / resources. But I’m not sure it even solves the problem anyway.

Yeah, I’ll see if I can do this tomorrow.

Thanks, that’s helpful! I’ll circle back on that a bit later in the week.

10

This has gotten much worse lately, probably because our Community - SocialHub category follows categories on other servers:

@angus Could you un-follow those categories?

1 Like
11

I’ve already terminated the connection two days ago.

1 Like
12

Thank you :slight_smile:

The “Latest” feed still shows federated posts, but they seem to be coming via other connections.

13

Yes it looks like the mirroring to Connected Places is still active although it captures replies as well, so everything Laurens replies to gets added to that category as well.