Comment by BrenBarn

Yeah, I think this is true and it's unfortunate. I think it's one of the areas where what seemed like the plan for Matrix hasn't totally worked out in practice. And it's even just the intricacy of managing it; some of the problems are deeper in the design. When I started using it several years ago, it seemed their vision was a lot of people running a lot of small servers. But the full-replication nature of the protocol, plus the resource demands of the server software, make that kind of impractical. Even tech-oriented people may shy away if they find out the database could grow to tens of gigabytes, and Synapse is not exactly light on RAM or CPU either.

Perhaps in the future if implementations improve some of this may get better, and it will become more feasible for small operators to run their own servers. But by that time it will be harder to build trust because too many people will have written it off as bloated or unstable. I think it would have been better to start lean and keep the system in more of a nerd niche until that process of evolution reached a later stage.