Right now, instances are the closest corollary to communities. Ideally, your instance is populated by people who are similar enough to you in enough ways that your conversations are rewarding, and your federated timeline more relevant than not.

But instances are administrative and technical obstacles, and discoverability is not good. There’s a potentially a lot of duplication of effort across instances, some of which reflects community values, some of which are probably shared by all users. (e.g., community content moderation versus global security)

Also, caching is a _thing_, if you want to grow a large instance.

So, obviously, just don’t do large instances at all, right? But then we’re back to the discoverability, administrative, and operational problems. (All of which can be pitched as features, if you look at them from the right angle).

Of course, these are all social graph problems at heart. Path—my all time favorite social network—was all ablut tiny communities of besties and/or families; LinkedIn is self-selecting professional relationships (effectively instance-based); Facebook and Instagram are basically just the whole of humanity, a few of which you’ve favorited; Google Plus (as well as I remember it), was very forward about allowing you to create different “circles,” but we all balked at the overhead of it all.

The closest one can get to the benefits of a true nomadic mastodon identity is to spin up a single-user instance. I think this is funny.

I think there’s a lot that be done to address the user-facing impact of these issues at the client level through thoughtful UX (especially multi-modal interfaces). Onboarding and discovery, in particular, are so bad out of the box that almost any improvements will be meaningful improvements.

I was pretty disappointed when @bass_rock told me that adding new and different types of timelines looks to be a pain. That seems like an obvious point for extensibility.