It is actually a better question than I initially thought.

the question actually challenges all sorts of assumptions we make about how the internet works. in this regards it reminds me of IPFS.

And while ipfs has some very cool tech the main thing it has problems with are names. ipfs has no names, which sucks. there is a name system ipns but it does not help much.

ipfs does a publish/subscribe subsystem which I find one of the most compelling parts. it is however considered experamental and inefficient.

this did not stop me from writing a distributed video streaming system using it. which turns out to suck, no names, and, even on my local network I was getting ~ 2 minutes of lag, no one would put up with that for streaming.

https://nl1.outband.net/fossil/ipfs_stream/doc/tip/readme.md

> That seems like a very, very poor choice.

By all means, enlighten us to the better one. It seems like they made a tradeoff here, and one that’s most in line with keeping everything distributed.

I don't know that I'd agree that it's a "very, very poor choice" but I can see how it would prevent Mastodon from seeing more widespread adoption. If widespread adoption isn't their goal, then it really doesn't matter though, and as you said, they made a tradeoff. I can definitely see the benefits of both ways of handling users.