Many register on to join the . It is perceived as the easy choice allowing wider reach.

People tend to overthink the choice of their home instance, but there's some truth to the worries. From a small instance, it might be harder to reach an audience scattered on many other small instances. Often you rely on a pre-existing network of contacts, either on or off-fediverse, to boost or follow you. ( can help.)

· · Web · 1 · 3 · 6

It doesn't have to be that way. relays can be used to make sure that all users in participating instances can reach each other's federated timeline. For visibility it's the equivalent of a mega-instance.

Individual instances can still moderate independently, but there are trade-offs. Seeing more content from other instances increases the chances you won't like something; trends become more homogeneous (and may require more curation if you don't like that).

The most commonly used relay is and there are several known instances .

So far, the only relay I found to be working transparently and reliably is by @xaphanpl. (Thanks!)

If you're a , consider joining or running your own (for the instances closest to you).

@gubi Go to your Mastodon instance's /admin/relays page and add the relay's /inbox URL, e.g.

@nemobis @mark

There’s an invite-only relay run by @aussocialadmin specifically to connect Australian instances. is connected, it seems to work well.

Having said that, I have Trends turned off on because it’s just not very useful for us and I don’t have time to whitelist things. Originally pretty much every trending hashtag was porn related, which was exceedingly boring.

@hugh @mark @aussocialadmin Interesting, I didn't know about the Australian experience! Very nice to hear.

I think I've never seen any porn hashtags but I can imagine how some instances can overwhelm trends that way...


It’s probably no longer an issue but when the functionality first was introduced I think there may have been a bit of deliberate gaming of it. My instance is much smaller than the others on the relay, which are open-registration.

@mark @aussocialadmin

@nemobis @xaphanpl

Interesting. I added the repeater, but then I suspended it because the timeline was out of control, too dispersive, with non-compliant posts.

@pierobosio Thanks for trying. I understand. Did you feel a few instances in particular were problematic (in which case it could be enough to silence/suspend them)?


No, I don't like to suspend instances. The main problem is that most of the posts are in languages other than English and Italian.

@pierobosio Ah! And they were incorrectly marked as being in English, so that filtering them in one's preferences doesn't work? Hopefully that will get better since per-toot language selection is now possible in v3.5.3 .

Yes, however, I want to keep my instance open to each language, if i need to follow a German account, for example.


Well, That's good.
Europe has been able to make the single euro coin, but has not yet managed to make a single language.

@nemobis An important thing to know for admins is that on a small instance it significantly affect resource usage since you're suddenly federating much more.

@nemobis And you probably won't gain anything in relation to "instances closest to you", since you're already federating with them.

@jaranta Dunno. The mere fact two instances know about each other doesn't mean that they know everything about each other. If nobody on my instance interacted with an account on yours, I will not find it by a string search. If nobody follows it, I won't see (almost) any posts when I check the account locally (or search by hashtag).

On there are two posts from the past 3 days, on only one because has no followers.

@nemobis Hmm. I thought federation was handled on the server level, but this sounds more like its a account level thing. I guess I was misinformed.

@jaranta I'm not quite sure I fully understand it either. I've not studied the source code much. by @blog got me started, the rest I learnt by observation on /admin/accounts, /admin/instances, /tags.

@jaranta Indeed. Relays had a significant impact on the sidekiq queue when it was on a shared server (especially if the relay happened to contain Poast). Now that it's on a decently sized VM, I barely notice a difference.

@jaranta When I looked into the slow jobs, often it was due to half-broken instances which caused my instance to wait forever on HTTP requests to cache things. A relay can help by excluding such instances automatically until they recover. seems to do so. (Not sure whether this feature is in master upstream.)

Sign in to participate in the conversation
La Quadrature du Net - Mastodon - Media Fédéré est un serveur Mastodon francophone, géré par La Quadrature du Net.