Repository navigation
Replies: 8 comments 10 replies
|
I think one of the big questions relative to cross-instance discovery is : Do we use AP for that or do we use Canimus/Chorus (however it's named) or a blend of both, and if so what's the best solution to solve what problem ? Part of Fairplayer's work seems to be oriented towards precisely helping "pull together instances", and I wonder if it wouldn't save work to think of a Mirlo instance as "integrating some Fairplayer modules / core logic"? |
|
This feels like it gets at some existential stuff honestly and I wonder if it relates a bit to some of the tensions that @timglorioso has talked about. I think for the sake of building an all in one tool for artists and listeners, it makes sense to do what you propose @tonigf and that is for a listener to follow artists on an other instance and get their posts and albums in a feed. What I'm feeling tension about is that maybe this is actually where separating the artist experience from the listener experience starts being too divergent. I'm not sure that this so complex that we need to seriously pause on this, but that is something that gives me pause. But this is also related to the Profile convo we've had a bit of back and forth around. (Should the Artist model become renamed a Profile model, and then should a listener profile inherit from the Profile model). Edit: also going to add this discussion https://socialhub.activitypub.rocks/t/fep-be68-audio-objects/8614/18 from @petitminion |
|
I think using artist makes sense as both the follower/followee. Like Si mentioned, in the data model we're leaning towards renaming "artist" to "profile" so that a profile encompasses effectively all artist and listener features. For example, listeners could make music blogs or podcasts, which are pretty similar to posts and releases that artists make today. That also makes it easier to have a single Mirlo account that can access multiple profiles, whether you're in a band and run your own label, or in multiple bands, or you make music and also want a separate personal profile to share thoughts on what you're listening to recently. |
|
Thanks all, The Profile model answers my second question. If both the follower and the followed are Profiles, then "who follows" is no longer a hard choice, the consuming side would just sit on Profile. @simonv3, about the artist/listener tension: I think Profile is part of the answer. Following + a feed becomes one feature that both use, so it brings them together instead of splitting them. But I get your point about not rushing. One thing worth saying: Canimus is already part of this. Fairplayer reads Mirlo through the Canimus feed today. So it helps to see two tools for two jobs:
So a lot already exists, just in separate pieces: the catalog works, the social publishing works, and the real gap is the AP consuming side. That leaves two open questions:
|
|
Thanks all. A few things from my side. On the rename: yes, I'd like to help with the Artist → Profile rename. Building on the existing WIP branch sounds like the right way to start. On Canimus vs AP: I see the appeal of going AP-native. My only hesitation is that Canimus already works and connects Mirlo to Fairplayer today, while the AP audio side is still being defined, so maybe we keep both for now and converge over time rather than switching all at once. Curious what others think. On the Channel model, my real question is: For a single artist who manages their own profile, I don't think we need Channel, Profile/Label seems enough, where Channel might earn its place is a different case we'd really like to support well: letting a label manage the profiles of several artists, in a simple, first-class way. So:
|
|
_
_ I think I'm lost as to what we mean by "manage" here. I understand quite easily what "managing a Mirlo artist account" is, but what does "managing over federation" implies in terms of features envisioned? |
|
I'm investigating bandwagon and it seems they use an approach that is more close to what you want : Funkwhale is more complex (it has an Upload object and channels) because it allow to privately share audio content, it allows deduplication of audio metadata, various audio files for the same track etc. If you want to implement federated user features like :
In any case it would be nice if you could coordinate with Bandwagon and if you could extend the FEP audio proposition with the choices you make. So every other project has clear documentation and we can all federate together ! :) |
|
FWIW I wrote some installation scripts and docs for getting things more set up on your own instance |
Uh oh!
There was an error while loading. Please reload this page.
Hi all! I'm Toni. I've been contributing to Fairplayer, and lately I set up Mirlo locally to understand its ActivityPub. Si suggested I move this here so more people can weigh in.
A bit of context on where this comes from: at La Instrumental we're quite interested in deploying our own Mirlo instance. That's what got me looking into how instances could relate to each other.
Reading the code (
src/activityPub/federation.ts), the 'publishing' side of ActivityPub is solid: an artist can be followed from the Fediverse, and their posts are delivered to followers. The artist consent is there too (thefederatedStreamingflag). What doesn't exist is the 'consuming' side. There's no Follow initiation, no.on(Accept)/.on(Create)listeners, and no data model for "remote artists this instance follows".Right now Mirlo is a 'publisher': an instance exposes its artists, and others (the Fediverse, Fairplayer) can read them. It doesn't 'consume', it can't follow a remote artist and receive their releases.
So here's what I want to raise: the moment we have several Mirlo instances and we want them to 'federate' with each other, Mirlo has to stop being only a publisher and become a 'consumer' too. For instance A to show artists from instance B, A has to consume B.
Open questions
Very open to all thoughts.
All reactions