[Cryptography] DAG vs Blockchain

[Howard Chu]

Ray Dillinger wrote:
> Now, everything I've seen so far has Bob either downloading every block
> of every branch hunting for that transaction, or blindly (without proof)
> accepting the assertion of other nodes that no such transaction exists.
> Bearing in mind that those peer nodes have the same problem as Bob in
> terms of not having complete information.
> 
> I don't believe blindly trusting peer nodes also in possession of
> incomplete information, without proof, is an adequate guarantee for a
> cryptocurrency application. And if Bob has to have every block of every
> branch down to the current moment, then what's the scalability advantage
> over a block chain?

Perhaps DAG is too vague for this application. You could implement something 
more specific - a tree (or a TRIE, actually). E.g., given output IDs of XX 
bytes, you might have a root/genesis block with 256 children. An output with 
ID beginning with byte xx would only be stored on a chain descending from the 
xx'th child of the genesis block, and you'd use DHTs to partition the 256 
branches across the entire P2P network. And below those first 256 children, 
you could use the 2nd byte of the ID to subdivide further, and distribute 
further. Then only small fractions of the network are responsible for 
propagating and storing small fractions of the transaction traffic.

The problem of course is nodes all have incomplete information, and whenever a 
verification is needed, most nodes will have to find some other network peer 
to assist, instead of having a local chain of authoritative info. And the 
question is will this additional volume of verification queries exceed the 
savings from the reduced propagation traffic.

-- 
   -- Howard Chu
   CTO, Symas Corp.           http://www.symas.com
   Director, Highland Sun     http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP  http://www.openldap.org/project/