The reason that small servers struggle today is not because their workers are written mainly in Python rather than Rust. Hell, they typically don’t run workers at all. The slowness comes from state res being slow algorithmically; state storage being slow and inefficient; wasting time trying to talk to dead servers; no support for “thin nodes” but always doing full mesh federation; the fact faster room joins never got finished; etc etc.
ALL of this would get fixed on FOSS Synapse. The point of Synapse Pro is to try to get $ to actually fund that work. Only massive-scalability stuff is in scope for Synapse Pro: all other features, perf optimisations, maintenance, security work etc will land in FOSS Synapse as it has for the last 10 years. Assuming that this gambit works and there’s any $ to fund it, of course.
Thanks Matthew for reassuring that FOSS Synapse isn't getting left behind. I really do want Matrix to succeed and get the funding it really needs. I'm aware it's been a long-standing issue.
My voice is mostly coming from the perspective of someone who has tried repeatedly to bring friends and family to Matrix but every time the experience is subpar. It's frustrating because most of the problems they experienced (including performance) have been known about and complained about for a long time.
Maybe the messaging in the article was a little off. The "attack" it makes on FOSS Synapse reads to me like it will never get the fixes it needs to make using Matrix more approachable to new users.
The reason that small servers struggle today is not because their workers are written mainly in Python rather than Rust. Hell, they typically don’t run workers at all. The slowness comes from state res being slow algorithmically; state storage being slow and inefficient; wasting time trying to talk to dead servers; no support for “thin nodes” but always doing full mesh federation; the fact faster room joins never got finished; etc etc.
ALL of this would get fixed on FOSS Synapse. The point of Synapse Pro is to try to get $ to actually fund that work. Only massive-scalability stuff is in scope for Synapse Pro: all other features, perf optimisations, maintenance, security work etc will land in FOSS Synapse as it has for the last 10 years. Assuming that this gambit works and there’s any $ to fund it, of course.