| 16 Mar 2024 |
 @fractivore:cyberia.club | In reply to @gavinrogers:matrix.org
in fact, it could potentially lead to a very innovative homeserver in the nixos community, the principles of which could be imported into the matrix ecosystem at large True, it's just a really huge ask, to make a new protocol. I know at least one person who is working on a novel federated realtime chat protocol, but it's like a huge project. Matrix is already pretty far along. | 20:22:11 |
 Gavin R | as someone said, why move it to #offtopic:matrix.org when we've just been moved from the (new) main room | 20:22:13 |
| Gavin R | matrix needs to make the time to consider these protocol level adjustments. saying it's too late now would be a huge mistake | 20:22:49 |
| @fractivore:cyberia.club | In reply to @gavinrogers:matrix.org
matrix needs to make the time to consider these protocol level adjustments. saying it's too late now would be a huge mistake I see what you're saying, ur rite ur rite | 20:23:24 |
| Gavin R | novel ideas are there for the taking. even some of the ipfs concepts need to be brought on board or the victim of success thing will be real. as someone said, there are a lot of users and potentially soon that will multiply | 20:23:34 |
| Gavin R | to start with, federation works best when the members can validate what code the other members are running. they don't have to be identical, they just have to be (mathematically) provable. this is something we learned in the ceptr (later holochain) project | 20:26:16 |
| Gavin R | guix and nix are the only current projects that can provide that to federated ecosystems (there might be lesser known projects i'm not aware of and i discount holochain because it ... well i'm not going to tell the story but it's a write-off) | 20:27:20 |
| @fractivore:cyberia.club | Personally, my biggest focus when it comes to improving matrix is working on threads. I want to do it, but I feel like I need to learn more about the protocol to really tackle it. If anybody wants to help, maybe we could form a sort of "task force" for threads. | 20:28:10 |
| Gavin R | they seem to work it's just a lack of uptake, people don't interact with the UX | 20:29:09 |
| Gavin R | threads are very important though - but i think the would be best implimented once some kind of proper graph relationship is acknowledged | 20:29:49 |
| @fractivore:cyberia.club | In reply to @gavinrogers:matrix.org
they seem to work it's just a lack of uptake, people don't interact with the UX They're totally unimplemented on a lot of common clients, like fluffychat which uses the "nested replies" fallback. I think a good start would be to get them implemented at all on common clients which don't support them at all yet. | 20:30:48 |
| @fractivore:cyberia.club | In reply to @gavinrogers:matrix.org
threads are very important though - but i think the would be best implimented once some kind of proper graph relationship is acknowledged Can you elaborate more on that "proper graph relationship acknowledgement"? I'm not following. | 20:32:01 |
| Gavin R | when the room split earlier, user jesopo described it quite well | 20:33:13 |
| Gavin R | it's a natural result of having messages exist in a federated world | 20:33:33 |
| Gavin R | the problem is that it's not officially acknowledged. we acknowledge the same problem in our software dependancies and the ceptr project was one place that saw the relationship to message systems | 20:34:51 |
| Gavin R | at least i and one other guy did | 20:35:02 |
| Gavin R | i mean you can solve this problem with Token Ring and etc | 20:36:21 |
| Gavin R | basically make sure that no one gets out of sync | 20:36:32 |
| Gavin R | Token Ring networking didn't scale super well, it was ok for some LANs. i think we have the knowledgebase within the nixos community to build something that works on a DAG | 20:37:17 |
| Gavin R | there is a single starting point that everything points to - basically whatever user writes the first "hello world" (quite literally, as a message) | 20:38:02 |
| Gavin R | everything else.... should point to something that points to that. if it doesn't, it doesn't belong. and we can make tools which allow the mathematics to do all the work for us | 20:38:44 |
| Gavin R | in fact you can apply set theory to groups of messages | 20:38:58 |
| Gavin R | that could prevent a lot of bandwidth wastage of which the matrix protocol is very guilty, let's be honest | 20:39:52 |
 K900 | Matrix room state is a DAG | 20:40:18 |
| Gavin R | there are optimisations now, it's not as expensive to run a homeserver as it used to be, but those optimisations are not elegant | 20:40:27 |
| K900 | You're literally describing how it already works | 20:40:31 |
| Gavin R | i know i'm just saying the docs and the code dont work from that conceptualisation | 20:41:17 |
| K900 | The problem isn't the DAG, the problem is consistency | 20:41:10 |
| K900 | I don't know what docs you mean, but I've looked at state resolution code in both Synapse and Ruma, and both are very much a DAG | 20:42:27 |
| Gavin R | agreed! and i think that comes because the people working on the implimentation(s) aren't aware of what they are working on | 20:42:27 |
