4 ms·

> This one is way off. Synapse, the "one complete server" you must be referring to, is going to be deprecated. The Matrix Foundation team is working on Dendrite

> This one is way off. Synapse, the "one complete server" you must be referring to, is going to be deprecated. The Matrix Foundation team is working on Dendrite to replace it. I hugely disagree. I have tried to setup a "homeserver". Synapse is huge beast, complex, CPU and memory-hungry, hard to setup (no, a Docker or VM instance does not mean easy to setup). Dendrite is not ready and has been "not ready" for years, not to mention it is also huge and hard to setup, and from the same vendor, so it allays none of the fears. Construct is the _only_ one of all the 3rd party servers which I have found to be actually in a fully usable and federable state (the rest being all in different states of abandonment and uselessness), created by an author who is _openly_ critical of Matrix-the-company and their policy regarding 3rd party servers (incl. how most of the protocol is just defined by bug-by-bug Synapse compatibility and how the Matrix webpage to this day has no 3rd party servers list -- but happy to feature 3rd party paid-hosting list) Compare this with Jabber. I install in seconds any of the _multiple_ servers that come with Debian stable. They are by now years old, but they still can connect with _everyone_ on the Jabber network. In Matrix, if I run a 2 year old Synapse, it will be like death-by-ostracization; same if I run today's Conduit.

moshmosh 5y ago
Similar story. A couple years ago I tried to set up Synapse. I've set up and administered IRC before. I've twice gone from nothing to having an XMPP server (different daemon each time) working and hooked into web services and custom auth serving production traffic. And that's just the chat-related work I've done. I kind of know what I'm doing, at least. After a couple hours I gave up on Synapse, having accomplished nothing and only being more confused than when I started out, as far as any path to a deployed config I'd be comfortable with. That they sell hosting, and it was such an incredible pain in the ass to run sanely on one's own, sets off some alarm bells.
vlmutolo 5y ago
I don't think anything you wrote contradicts what I said in the parent comment. In the section you quoted, I only meant to say that other homeserver projects are definitely not dead. There are a bunch that are actively developed and improving. It's also true that the new homeservers aren't yet ready for prime time. But they're coming along really fast. It's true that homeservers by third parties had a bit of a rough time in the beginning due to finding "bugs" in the spec that no one noticed before because there had previously only been a single homeserver implementation. That situation has been improving though. I've seen Conduit development result in several different patches submitted to the upstream Matrix spec.
- AshamedCaptain 5y ago
  > It's also true that the new homeservers aren't yet ready for prime time. But they're coming along really fast. I have seen way too many being abandoned to believe that. In fact last time I researched I put Conduit in the abandoned list, though apparently I was wrong. Matrix is just a too complex protocol, and way too focused on one and only one server implementation. Which is ironic considering probably every other competitor protocol (save for perhaps IRC itself) is even more complex, but at least they were not that complex when they initially showed up -- so other people could actually start implementations that survived and growed along.
  - vlmutolo 5y ago
    I asked the primary author from conduit, as a third party implementating a server, if he thought the protocol was too complex. He said no, which actually surprised me. I'm somewhat of an layperson on this topic in that I haven't actually ever tried to work with the protocol on a low level. But my take on it is that if they can slow down the pace of Matrix evolution in the next couple years, the protocol will be in a good state. It's complex, but it's providing extremely complex features. But if they keep going at this pace for too long, third party implementations may become less feasible.

moshmosh 5y ago

Similar story. A couple years ago I tried to set up Synapse. I've set up and administered IRC before. I've twice gone from nothing to having an XMPP server (different daemon each time) working and hooked into web services and custom auth serving production traffic. And that's just the chat-related work I've done. I kind of know what I'm doing, at least. After a couple hours I gave up on Synapse, having accomplished nothing and only being more confused than when I started out, as far as any path to a deployed config I'd be comfortable with. That they sell hosting, and it was such an incredible pain in the ass to run sanely on one's own, sets off some alarm bells.

vlmutolo 5y ago

I don't think anything you wrote contradicts what I said in the parent comment. In the section you quoted, I only meant to say that other homeserver projects are definitely not dead. There are a bunch that are actively developed and improving. It's also true that the new homeservers aren't yet ready for prime time. But they're coming along really fast. It's true that homeservers by third parties had a bit of a rough time in the beginning due to finding "bugs" in the spec that no one noticed before because there had previously only been a single homeserver implementation. That situation has been improving though. I've seen Conduit development result in several different patches submitted to the upstream Matrix spec.

AshamedCaptain 5y ago
> It's also true that the new homeservers aren't yet ready for prime time. But they're coming along really fast. I have seen way too many being abandoned to believe that. In fact last time I researched I put Conduit in the abandoned list, though apparently I was wrong. Matrix is just a too complex protocol, and way too focused on one and only one server implementation. Which is ironic considering probably every other competitor protocol (save for perhaps IRC itself) is even more complex, but at least they were not that complex when they initially showed up -- so other people could actually start implementations that survived and growed along.
- vlmutolo 5y ago
  I asked the primary author from conduit, as a third party implementating a server, if he thought the protocol was too complex. He said no, which actually surprised me. I'm somewhat of an layperson on this topic in that I haven't actually ever tried to work with the protocol on a low level. But my take on it is that if they can slow down the pace of Matrix evolution in the next couple years, the protocol will be in a good state. It's complex, but it's providing extremely complex features. But if they keep going at this pace for too long, third party implementations may become less feasible.