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.
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.
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.