Looks good! For me the gossip protocol is probably less interesting because in SAFE the peer-to-peer layer is abstracted away from the application layer; an app only sees a giant hash table of data. It’s of course possible to again implement a gossip protocol on top of that using the messaging system, but that’d be way too slow and expensive for most purposes I suppose.
From the perspective of the application layer I find it useful to look at SAFE as a reliable but dumb database. Dumb because the user agent can’t ask it any complex requests, it has to piece everything together itself using many basic requests.
As things stand now in Project Decorum’s design user agents rarely directly message (not talking about personal messages here) each other; they just publish interesting data (from their perspecitve) in such a way that it’s easy for others to find it if they (are likely to) have a common interest. The user agent has to be pro-active if it wants to know anything, no other agent is going to directly tell it what’s what.
I suppose the more fundamental difference here is the strict separation in SAFE between vault and client functionality. Advantages are that a client can make do with very poor upload bandwidth and maybe has better privacy. A disadvantage is that the network probably uses more bandwidth in total (vaults are middlemen here, a detour). We’ll have to see how it’ll all turn out in practice, the only thing I know for sure is that it’s going to be fascinating to see it all evolve!
I certainly appreciate the principles and design of scuttlebutt, it strongly resonates with what we’re trying to accomplish here.