By making only optimisations that primarily support massive servers proprietary (for now), the $ from Synapse Pro can hopefully then directly fund the features and maintenance of core FOSS Synapse, including core optimisation work which benefits everyone. Concrete examples include:
* State Res v3, which hopefully should have algorithmic improvements to perf
* State storage, addressing the long standing diskspace consumption problem for state snapshots
* Finishing faster room joins so they are actually fast
* Shared retry hints on federation, to stop wasting time trying to talk to dead servers
* Better than full mesh federation, so small servers can participate in big rooms without melting.
EDIT: to spell it out, as it’s been misunderstood: Fixing these issues will likely happen directly in FOSS Synapse (assuming there’s any $ to work on them at all).
Meanwhile, FOSS Synapse already has Rust as Python codepaths and has done for years, and that will only increase over time. Only stuff for making workers scale to handle hundreds of thousand concurrent users will remain proprietary for now.
Dendrite was meant to be the next big thing, but we (obviously) didn’t have enough $ to work on both it and Synapse. So we focused exclusively on Synapse as of end of Dec 2023, and Dendrite is currently progressing on a best-effort (ie not funded work) basis.
> It feels like element is often jumping on new things like big rewrites such as element X instead of improving the existing products.
Well, Dendrite was a big rewrite like that. Which is why we’ve ended up improving Synapse instead (both FOSS and the new rust workers).