Question for the #FediVerse , how do I make mastodon actually fetch replies to a post? Someone boosted a post onto my timeline, but my instance doesn't have any of the replies to it. If I open it in a new window, I see a bunch of replies.
Have to do all that extra work is a PITA, and that means if I want to interact w/ any of the replies, it's THAT much more work (copy/paste urls back to my instance).
One approach is to open the post in the UI of the instance.
The obvious other hack to perform would bet to do something similar to relays(e.g. https://relay.101010.pl/ -- services that basically boost everything they receive), so: observe all the posts that you receive, poll their lists of replies, and boost them. Having your instance's actor follow a bot that does so would prefetch those messages into the instance.
IOW, we are in the same boat in which mailing lists are with inability to retroactively add ccs. We can "solve" that with something morally equivalent to a bot that always gets ccd and then, when it notices the cc list expanding, resends previous messages to the new participants (and keeps doing so for new messages in split-off threads).
@robryk
Re: open the post in UI of instance is what I said when I open up a new window, as that is what mastodon does.
The thing is that mastodon HAS the URL for the post (otherwise I couldn't open it up) so why doesn't mastodon just simply query the server and say, hey give me all the replies to this post?
The difference with a mailing list is that there isn't a defined server protocol for fetching posts.
When would it do that? Note that even barring possiblity of races (given that everything works on message queues), you can have new replies to messages upthread (or cross-thread) from your message in that thread not be sent to you.
If you want to say "regularly", then this is a terrifying DoS vector for small instances and IMO it's a good principle not to initiate outgoing connections to instances other than as a result of user action.
I would dearly wish that ActivityPub instances had a concept of a "thread" that can be followed. (Note that all instances I've seen differ from e.g. Google+ or FB in that G+ or FB have such a concept _and the author of the initial message has ownership over the thread_. I don't know if that's a good model, but in my limited experience that model has somewhat more clearly delimited expectations of different participants wrt moderation of responses in threads. For better or worse, threads that can be followed would approximate that model.)
@robryk
When the user "clicks" on the message. Yes, there'll be a delay, but I'm not saying poll all servers for all messages ever displayed to the user for replies.
I only "expand/click" on a few messages a day, compared to the number I see in my TL.
If the protocol can't handle this, then is needs to fixed/replaced.
(Yes, already know of the ddos attack by posting urls and all the instances getting a preview, this should be lower weight than that.)