YSK: There's a protocol designed for decentralized networks. It is called Named Data Networking (NDN)
YSK: There's a protocol designed for decentralized networks. It is called Named Data Networking (NDN)
Instead of one-to-one connection it implements one-to-many. Theoretically it can significantly reduce costs of self-hosted instances, increasing speed and reliability.
Is anyone interested in applying this to fediverse, or is it only me?
@qnick you might also like https://reticulum.network/
Looks like it works only over IP, which is by design one-to-one. The point of NDN is to get rid of IP connections
@qnick it can work over IP but doesn't require it. A common use case is over LoRa links which don't have IP https://reticulum.network/hardware.html
This is an interesting alternative to TCP/IP. I’ll need to dig into it a bit to have any real opinions.
Saved for when I start my own instance, thanks
This seems neat. Networking isn't my main area, and I'm still reading about this stuff, so please forgive me if these are dumb questions, but this has raised my curiosity:
Ip does have multicasting, correct? Could multicasting accomplish some of the same one to many goals? Does NdN provide performance/ease of use benefits over multicasting?
From what I can tell, ndn can be implemented over ethernet (preferred) or over ip (mostly defeats the purpose). Is ndn over ethernet actually a thing that can happen now - ie will whatever routers are in the way between various things that want to talk to each other handle it correctly? If not, is this likely to happen in the future? If so, is ndn over ip worthwhile in the meantime, or is it better to wait?
It seems like cool technology, but a lot of what I've read suggests that it's still in developing stages, so it's unclear how you'd actually use it just now.
IP multicasting works when many clients request the same data at the same time, it makes sense only for realtile streams, and you need to implement caching by yourself.
In NDN caching is already included in the protocol and implemented in the forwarding daemon. It works similar to CDN. If all lemmy instances were connected to one NDN, the cache data would be spread across the entire network. So even when lemmy.world is offline, you'd be able to read it from the nearest (by ping) lemmy instance.
I stated that getting rid of IP is the goal of NDN, but it obviously cannot happen right away, because lemmy instances currently don't have any direct connection to each other. But it will eventually happen, so we better get prepared.