3 ms·

So should you assume a linear cost (in terms of latency) in the distance in the key space ? This looks like a weird tor overlay network, to be honest, with ever

So should you assume a linear cost (in terms of latency) in the distance in the key space ? This looks like a weird tor overlay network, to be honest, with everyone on a line… Also: is there some white paper / RFC explaining all the rational ? It seems matrix always put the implementation before the specification, which seems a bit weird (and might be why you already have like 5 version of objects such as rooms).

kitkat_new 5y ago
trying to implement it first seems sensible to me as you want to ensure you have a concept that works in practice before you put it into an official specification
- Arathorn 5y ago
  yeah, the whole point is that we demonstrate that stuff works before formalising it in the spec. Pinecone itself is pretty early (and keeps changing) though and hasn’t even got an informal spec yet. But we will soon! edit: meanwhile, https://spec.matrix.org/unstable/proposals/ https://spec.matrix.org/unstable/proposals/ is the index of all the spec changes which are in flight. and if you’re interested in understanding how the matrix spec evolves, https://m.youtube.com/watch?v=SFkZz60RRfc https://m.youtube.com/watch?v=SFkZz60RRfc is an hour of yours truly trying to explain the process and its rationale.
neilalexander 5y ago
Not exactly — the line topology is effectively a “minimum” topology, whereas the bulk of the actual routing knowledge is built up from keyspace neighbours searching for each other and creating virtual paths via intermediate nodes.