This is VERY bad news for users of small instances like myself.
I do understand the reasoning behind it, and I'm not sure there is a good alternative. But it sure feels a bit like a deja vu from the self hosted email days ...
@michael Agreed, this feels like a step backward. I'm curious if @Mastodon and/or @stux would be open to setting up a read-only relay service endpoint from the servers they manage. As 50-60% of posts come from users on their instances, such a move would offset the impact this breaking change
@michael The hashtag relays provided by FediBuzz were a great way to add content I'm interested in into my timelines. I guess as a user of a single-user instance, things are about to get more lonely 😔
@michael Interesting, I will check that out more later, glad to see there is a GitHub Action option that doesn't touch the infra, as I'm on managed hosting 😅 Thank you 😊
@christian yep. That Action was the inspiration for #FediFetcher ‘s Action and I’ve run GetMoarFediverse as Action for many months, and can confirm it works well.
@anders@ClearlyClaire my impressions has always been that she’s a kind and helpful person, and she’s personally helped me twice very patiently with hard to debug issues on my single user instance.
@michael@ClearlyClaire when I started work on combine.social and tried to have a conversation with her about it, she made it quite clear that they have no interest in making federation better for small instances.
And now we see steps which specifically make federation worse for us.
@michael the alternative is to sign the scraper up for an account on each instance it scrapes. they're easy to get on most of the big instances, and they wouldn't generate any more load than any other account that leaves the federated timeline page open all day but never posts.
@vyr Right. That seems better to me: signing up for an account means that TOS apply, and sites that don't want this kind of scraping don't have to approve the accounts. Of course it's a hassle for the scraper especially if they want to go beyond the top N instances, so I can see why there's resistance to it. And it seems like it's taking everybody by surprise so the communication hasn't been great -- these would have been great discussions to talk through before announcing the beta.
But also @michael I didn't mean to undercut your concerns, when I say "talk through" I mean with small instance admins as well as the fedibuzz folks. The solution as implemented seems likely to favor large instances over small ones, there might be ways to accomplish the goals with a more equitable solution. @vyr
@vyr yeah, I was discussing this with someone yesterday. Seems dodgy as …
When I said I’d get a token for each instance I thought they’d create their own accounts. Not that they’d get random people’s accounts…
@vyr the occurrence of sketchy instances doesn’t bother me as much - imho it should be the subscribing instances’ job to block what they want to block.
But deffo feeling sketchy now. Shame, as it was so useful.
Yeah it's certainly useful but there are multiple "yikes"es here. I'm not sure what the right answer is but this is not looking promising.
Taking a step back, one way to look at this is that the product implementation to date don't address a use case that's critical for small instances and cross-instance communities like Monsterdon.
So a hacky solution emerged -- relying on a security hole, and with some major shortcuts. As a result, small instances and cross-instance communities became reliant on the hacky solution.
Now, the security hole is being closed (good!), but the underlying use case still isn't getting addressed (not good). And, implementers of the hacky solution -- who knew about the upcoming change -- didn't alert the people relying on it, or start exploring an approach that would be less hacky and more secure; neither did the product team. Instead the implementers of the hacky solution waited until the shit hit the fan and then proposed an alternate hacky solution with even worse security implications. Not only that the attention from the discussion highlighted that at least one of the the shortcuts in the hacky solution (not forwarding edits or deletes) that people have been relying on has major privacy and safety implications).
Not good.
More positively, thought, it points out a threat I had missed in the draft Meta privacy threat model!
@jdp23 this is a general threat model, and Twitter had the same problem. i'd actually expect this from anything with an API:
apps/websites that claim to show you your top posts or who's blocking you or whatever, but are going to be asking for overbroad API tokens or straight up just asking for your password so they can also post spam or access stuff they shouldn't. many of them will probably have shoddy backends and leak credentials or data by accident, or share them with third parties on purpose.
i didn't have "websites that claim to show you top toots and also sell the data" on the bingo board, but i probably should have. anyway, here's "FeedSeer News":
> You grant us read only access to your Mastodon account, and we monitor the feed of status updates from the people you follow, just like your Mastodon client does.
for admins who might want to keep an eye on this one particular scraper, as useful as it is, the tokens are created with app name #FediBuzz and app website https://fedi.buzz/. (Mastodon autolinking is mangling the URL there, but it's the whole URL from https:// to trailing /).
note that they don't have to be. FediBuzz could just as easily pretend to be Toot! or Tusky or Pinafore. but it is currently politely identifying itself in a normal way.
@michael Hard disagree... I think that relays are a kind of snake oil. It just flushes your instance with a lot of noise, you don't care for. I did try it and it was just bloat for me and the other 10 users on my instance.
@michael@astro Just chiming in to say that FediBuzz has been really useful for my single user instance and not being able to use it would be a great loss for me.
I've found relay.fedi.buzz really important for my small instance, as none of the largest instances seem to be using any relays. Losing this will mean that my instance will again not be aware of majority of the posts on the Fediverse, and following topics or context of many of the conversations will again be very difficult.
@thisismissem mentions OAuth as a potential solution, but if I understand correctly, this will not be implemented at least in the next version.
@lari@astro@michael tbh, I don't think the way forwards (long-term) is to use the streaming API to scrape everything that's public, but instead to position the service as valuable enough that instance admins choose to add peerage to your service as a relay.
@thisismissem@astro@michael You're probably right about the Streaming API, but I don't think I or any other small instance admin can do much to make the large instances to join a relay. Most of the large ones aren't using any, and I haven't noticed that even those operated by @Gargron would be using them. When Streaming API is blocked, their feeds will be unavailable to my small instance.
This change will probably not have much effect to largest instances, but will affect negatively all the small ones which have been relying on services like fedi.buzz to be aware of what is happening in the #fediverse
@jay through each server admin giving access to each relay they want to participate in. However largish instances don’t usually do this, so fedi.buzz was really useful for small instances to follow relevant hashtags on large instances
@KI5SMN@michael
I use it. It fills my instance with content posts that larger instances, incl Mastodon .Social
have shut us out from by not opening their relays.
"Mastodon and many other ActivityPub-compatible services live and breathe decentralization. Yet it can get lonely on a small instance. Mastoadmins can bring the global buzz of toots into the Federated Timeline of their small server by following ActivityPub relays."
Read more: https://relay.fedi.buzz/
Isn't there a mechanism that lets users on a user's basis to stop their posts from being public or does that just set the robots.txt file for the user and people ignore that anyway?
Add comment