Marlin protocol

Peer-to-peer networks in virtually all decentralized systems are built to cater to an invariable pattern of communication: information produced by one node needs to be propagated to each and every node in the network. Such a one-to-many pattern is dramatically different from the predominant pattern seen in traditional server-client (or point-to-point) architectures. And unsurprisingly, much of the current point-to-point network protocols are unsuitable for decentralized systems.

Current decentralized systems are based on peer-to-peer communication via Gossip protocols. These systems operate by a sender broadcasting the message to its peers and then the peers recursively forwarding the message to their peers until the message has been received by the entire network.

While such protocols are well studied for reliable information distribution, they are routinely inefficient in terms of the 1) latency of propagation, 2) bandwidth consumption, and 3) network congestion. Furthermore, they lack incentive compatibility: for e.g. a Bitcoin miner does not have an incentive to forward the transactions it receives. In cases such as Zcash and Monero where anonymity and privacy are paramount, Gossip message forwarding also opens the door for an attacker observing the network to infer the IP/network addresses of message senders, thereby revealing private information.

This is a fundamental difference in SAFE. We have locality of knowledge with a global mechanism to identify where that information is located. More like nature where you know your own local and can tell folk about it, but the further from the local information then you know less about it, but you always know it is there.


Information is propagated to each node in a section,making it like a small network though,no?

Yes, each section is like a little village :slight_smile:

Although mentioned already this struck me as something decentralised networks do not do. Mesh networks don’t nor do many real decentralised networks, except maybe some (?nearly all?) blockchain networks.

And as David said SAFE is far from it. And as you said it is more localised to sections. But since sections are going to be very small to the whole, the statement is definitely wrong for SAFE and is actually wrong for the general large decentralised network concept. So has the white paper been based on a invalid concept or is it aimed for blockchain networks rather than decentralised networks???


It seems the first application is for blockchains,but they mention lots of other things. They’ve made a replacement for TCP, some routing protocol and other network stuff.