@nipos yes, you have to have a homeserver somewhere to speak Matrix. We could have embedded one into Gitter, or just have Synapse/Dendrite sitting alongside it. Using one sitting alongside is easier. Gitter isn't going to "completely die" at all; and if we fail to converge Element & Gitter's UI, then we'll keep the Gitter engine running. Meanwhile, anyone can continue to run their own Gitter, with their own homeserver is they want, whatever happens on Gitter.im.

@matrix I think you're missing my point.Yes,I can run my own Gitter but then I still need Synapse which has huge performance issues.With the announcement that you buy Gitter,I hoped that Gitter will become a alternative to Synapse with it's high-performance backend.And you said that Gitter the software will be replaced by Synapse and Element yourself so Gitter the software will die.Gitter the platform will stay alive but only as one of many Matrix instances.I really like Matrix and run three instances currently but I think more diversity in software would be good for the protocol.

@nipos Synapse's performance issues are way improved these days, as per https://matrix.org/blog/2020/11/03/how-we-fixed-synapses-scalability.

Meanwhile Dendrite is maturing nicely and outperforms the Gitter backend. It's very unclear what benefit there'd be to add a javascript homeserver implementation to Gitter's backend though - the most plausible approach would be to embed a Dendrite. At which point, why not just run a Dendrite (or Synapse) alongside it?

@matrix Yeah,Synapse works way better than it did in the beginning but still I'm not 100% happy.I think that for a protocol to be successfully adapted by many projects,it's important to have much software that supports it.There are many interesting Matrix servers in development but most are far away from production-ready.I'm currently doing experiments with Construct and I love it but there's too much stuff missing to replace Synapse yet.Why I want especially a Javascript homeserver: It's a language that's easy to learn and very flexible with the possibility to do changes without having to compile it.That lowers the barrier for contributing and enables the possibility of doing own projects by changing small parts step by step.Also I think that the existing Gitter software with Websocket connections instead of HTTP polling is by far more powerful and a lot faster.Another point is that Gitter has a lightweight client.Sure,it lacks some features and doesn't even have encryption but for public chats about open source projects,that's absolutely not neccessary.I love the loading speed and I appreciate the possibility to share a link to a chat and have anonymous users read it with nearly zero load time and without the requirement to sign-in.Sure,I can do the last thing with Gitter alongside a Synapse deployment too,but only for now.You said that your goal is to discontinue the Gitter client as soon as Element has all required functionalities.That would be the end for this option.I'm a big fan of both Matrix and Gitter but for different use-cases.Matrix is great for private chats and smaller closed groups but Gitter is superior for very large public chats about public projects.I can't even send public links to public Matrix groups most times because most Matrix servers just don't show the preview without the user being signed in.