Delete activities for a Note are only federated to followers, mentioned users and relays #190
Loading…
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
This means that notes federated to instances that do not follow these criteras (but are federated with the instance otherwise), e.g. via renotes, will never get deleted, as the Delete activity is never sent to them.
Ideally, the Delete activity should be federated to all known instances.
I'm not sure about 1) how this would impact performance, especially on "well federated" instances and 2) how this would impact queues clogging up because of suddenly trying to contact instances that have been dead for a while.
Of course (2) is just a symptom of a differen problem, but since that's not (yet) fixed...
I guess implementation wise we should go by this paragraph from the AP spec:
Addendum: it seems deletes for DMs do not federate from FoundKey.
While this is a MAY, I think that most of our users expect deletes to, well, do a best-effort to delete the post across the network.
If you can please check that case also with the linked PR. This might have been caused by the instance panicking when trying to deliver to followers which are not allowed to see the post. Delivering to followers should now also only happen in the appropriate cases, but the check was missing before.
One could look at how Mastodon handles it (or Pleroma, or others), mastosoc seems to be doing fine on that regard, though. Pleroma isn't a shining example in performance or code quality, so that can be set aside somewhat (but it still manages fine).
I believe Misskey/FoundKey regularly fetch nodeinfos and keep information as to which instances are up-to-date, in database. Re-using that information shouldn't be a problem.
As far as I saw there is nothing in the vein of "update instance info in a regular time interval", more like "whenever a post federates from/to an instance".
If I recall correctly, "automatically stop federating with dead instances" has been requested multiple times, which is what I was alluding to here and is a separate problem in my opinion. (Although I don't think there is an issue for it yet?)
Reopening this as it seems the performance hit is very big. We have to reconsider if this is a good idea and if so how the performance could be improved.