p2p network topology, why even bother?

re: %1o2VHqx...

Still not sure why [the secure scuttlebutt] network is needed (no offense to those who have friendships & collaboration here), except for the chance to build-test a p2p system with no servers. A bit like Beaker Browser?

~matt the motivation for me is to stop needing servers and intermediaries in general. sometimes intermediaries are necessary, but i like the idea of creating those systems in an ad-hoc manner using distributed (rather than polycentric) building blocks. it also fits better with the goals of sunshine garden's walkaway project, which cannot guarantee stable internetworking the way that on-grid internet can.

I look at Manyverse (when my phone can handle it) as what I always wanted (since I knew what servers were), a server in my pocket.

My first internet service provider in the early 1990's gave everybody the option of running a server, so I did, so I had a server on the same computer I worked on. SSB is still like that.

~xj9 I think that is the same as not needing servers, no?

~bob yes and no :)

part of it is a question of topology. federated networks are p2p if you squint at them the right way. they can also be viewed as centralized systems if you squint another way. this is why i tend to refer to "federation" as polycentric instead of decentralized. in practice a user is always a leaf node that has to use some other node to access network resources.



federation basically mirrors the physical topology of the internet, which isn't necessarily a good thing. you have to do everything through a remote server for a variety of reasons, but they are all consequences of the status quo. the federated / polycentric strategy works, but only if you ignore the reality that:

something like manyverse on the other hand is:

there's technically a server running here in both cases, but the admin overhead is much, much lower for the p2p network client. web-based federation systems are very complex and have a ton of moving parts, to the point where people are starting hosting companies and cooperatives to take care of it for you. i see this as a huge red flag.

i think i can agree that everyone running a server is almost the same as not needing servers, but there are important considerations that make or break this that aren't only related to ISP behavior.

~xj9 Yes. The problem with [distributed network diagram] ( C ) is that it fails to show that some nodes have a very large number of connections, and some have very few. I used to think that these were distributed in a [scale-free] manner, following a [Zipf] distribution, but its actually square root of that. Here are two examples: gene-gene-gene interactions (triangles, and also gene-protein-pathway pentagons) and also wikipedia readership distribution.

gene-gene-gene interactions

wikipedia readership distribution

~xj9 I think that the P2P network actually might be represented this way . .



bcos it actually may (typically will) exist in multiple dimensions, with nodes connected to each other thro multiple media. So, the graph is multidimensional. The 'plane' of the network as originally shown, 'wraps around' on to itself in multiple ways. Yes?

~Mike_Hales, ~Linas it isn't supposed to be a perfectly accurate model. just an illustration of the difference between centralized, polycentric, and distributed networks. the actual distribution can differ a lot, but the topology is still relevant to how nodes communicate. the exact details of centralized and polycentric networks are also not exactly expressed here. the goal was to point out certain topological differences, but i appreciate the interest in making the models more accurate! i think this chart is good for introducing the idea, but it would be neat to have more detailed and correctly distributed models for a more detailed discussion on the topology of a given network type.

all together now!