[Cryptography] (was credacash)
Howard Chu
hyc at symas.com
Sun Jul 2 07:14:11 EDT 2017
James A. Donald wrote:
> There is a lot of really serious money going into crypto currencies. A lot of
> very rich people are hedging against the collapse of the US$ and the US
> empire, and they rather suspect that it will be replaced by a crypto currency
> rather than the gold standard.
>
> But bitcoin is hitting its scaling limits hard. Maybe they will solve this,
> and become the one world currency, but at the moment we are instead seeing a
> whole bunch of monkeys flinging crap at each other as each seeks to shift
> scaling costs somewhere else.
Bitcoin's been at its scaling limit for over 2 years. Aside from that, Bitcoin
lacks fungibility, which by definition means it is *not* a currency.
> A lot of altcoins have appeared, many of them get rich quick scams. They
> generally claim to have solved the many problems bitcoin suffers from. Last
> time I checked, this was not exactly true.
Monero solves most of Bitcoin's problems - fungibility, dynamic blocksize
being the two major ones. But yes it's still a blockchain, which is still
inherently unscalable. As I see it, the only way forward, while still
maintaining the distributed, trustless properties of the blockchain, is to
partition (shard) the data. Without having worked out all the details yet, I
believe it can be made to scale by structuring the blocks as a TRIE, which can
grow additional levels as needed. Branching within the TRIE could be as simple
as indexing off a particular byte of a block address. Or it could be based on
a consistent hashing algorithm.
If you can successfully, fairly, distribute responsibility for block storage
across the network, you should also be able to reduce overall bandwidth
consumed by block propagation.
As an easier to implement option, you partition based on block number instead
of block address. In this case, every node still needs a complete list of
block addresses (in order to map addresses to number), but given N branches of
the TRIE they only need to store 1/N of the actual blocks. Then obviously,
only one particular subset 1/N of the network needs to participate in
transaction propagation at a time, also.
Assuming storage requirements are thus out of the way, the only limit to
transaction rate is network speed / transaction size.
--
-- Howard Chu
CTO, Symas Corp. http://www.symas.com
Director, Highland Sun http://highlandsun.com/hyc/
Chief Architect, OpenLDAP http://www.openldap.org/project/
More information about the cryptography
mailing list