Comment by jchw

I honestly don't blame them that Dendrite didn't work out, I blame them for leaving us all stranded without any sort of migration path. I don't personally care if my homeserver is written in Python or Go or Rust, at least not very much. Rust may very well be a better choice than Go in the long run. I only chose Dendrite because when I chose it I felt like it was in better shape than Conduit (which at the time wasn't forked off, wasn't well-supported by app services either, and didn't handle federated presence) and lighter weight than Synapse.

Is my homeserver even counted in the 369 active Dendrite servers anyway? I mean, I guess it definitely isn't in that count since it is now an active Synapse server instead, as of late November. However for what it's worth, I don't think I actually connect to Matrix.org at all. Maybe it still counts if you're in a room with Matrix.org users?

But assuming it is basically just 369 Dendrite servers, well, they're all stuck there. How hard could it have been to provide a migration path? I dunno. My attempt is definitely incomplete and possibly quite ill-advised. I truthfully don't understand Dendrite or Synapse well enough to design a migration tool that works correctly. Did I calculate forward extremities correctly? Maybe. Are my state groups actually all correct? Probably not. How the hell did I end up with rooms that have multiple auth events? I have no idea, I just wound up destroying the corrupted rooms I found. I can only imagine that whoever developed the Dendrite database structure also had knowledge of how the Synapse database structure works and would've been able to do all of this in many less hours than I was.

After all is said and done I'm sure the balance sheets will look much better but the damage done to the community side of things will linger. I won't feel bad if Element fails and Matrix winds up replaced by a different standard later.