search

dominictarr / scuttlebutt

peer-to-peer replicatable data structure
Other
1.32k stars 66 forks source link

Does scuttlebutt/events omit events? #36

Closed rkusa closed 10 years ago

rkusa commented 10 years ago

If I am reading the paper (Efficient Reconciliation and Flow Control for Anti-Entropy Protocols) right, the history of a certain key is not guaranteed to sync completely, i.e., in some edge-cases some old values/versions are omitted.

E.g. Paper p. 3 figure 1 and

A remarkable property of Scuttlebutt is that the gossip ex- change between two participants is not designed to eliminate all differences between the two participants, even if there is sufficient room in the gossip messages.

Have you made some adjustments to the scuttlebutt approach to guarantee full-history sync, does the scuttlebutt/events omit events, or do I get the paper wrong?

Just want to now, what to expect from scuttlebutt/events :wink:

Thanks

dominictarr commented 10 years ago

Well I think the heart of the protocol is really about log replication. it's the handshake that is the important thing. When they talk about 'old values' that is in a layer on top of that.

The way the use it, they forget information that isn't relavant - which is how scuttlebutt/model works, but scuttlebutt/events doesn't forget anything.

rkusa commented 10 years ago

Ok, if I get it right, you do a full-history sync instead of just syncing the most recent values, which makes scuttlebutt/events work.

Thanks for the info