You are viewing a single comment's thread from:
RE: Doing the impossible: decentralised fuzzy search on a DHT
Did you do any testing of this method? In particular, what if someone spams some searches on some popular 'shards'? Don't you think it would introduce a point of failure?
Why wouldn't anyone have done it by now, for something like BitTorrent for example then?
Can you code something like that?
Each peer in the network connects to a random set of other peers and thus has a unique view of the network. When any particular node runs a search with its currently connected peers, that is only a small subset.
In other words, each shard is replicated and mirrored on multiple nodes, because each lookup of a word returns multiple peers that have data about that word.
So this solution would waste a bunch of storage space.
Doesn't exact search on DHT usually work with bounces? Nodes relay the search (to less and less nodes depending on the number of already made bounces), and if a node with a positive result isn't connected to the node that originally sent the search, then connect to it. I saw a few DHTs using this technique succesfully. Clearly doing the same for fuzzy searching isn't possible as there is not positive (1) or negative (0) results, but also anything in the middle.
Seriously man, I'm very interested if your solution could work, and I'd rather see schematics or code rather than text. It would be much clearer to me.
I have some personal open source code of a DHT using WebRTC as a transport where I implemented exact search this way. I tested it and it's fine. I never really thought of any other way.
https://github.com/skzap/waka2/blob/master/peer.js
Can you code my friend?
I'll work on some actual code later today or tomorrow to demonstrate how it could work.