@ariadne@treehouse.systems i've looked at a few of these and the biggest problem i've seen is respecting the individual nodes' resources.

there's some inherent overhead added upon the baseline of a completely centralized system of course, but a lot of projects seem to not factor exactly /how/ much is added into the design.

see: my synapse DB was ~27GB a week ago after existing for ~2 years. my entire znc data dir is 2GB, and it ran for maybe ~3.5 years.

e.g. the CEK key is nice in theory, but if you look at matrix getting it working well seems to be very painful. and similarly, if OCAPs become an actual programmatic construct that could leak into bandwidth/disk space.