ZEN protocol

Asher Manning, [30.12.17 13:05]

'This is correct. We haven't implemented a sharding solution yet, however our throughput should be much better than single-threaded smart contracting platforms. Sharding is much easier if you don't have any shared state, and I don't think it would be particularly difficult to extend Zen Protocol with sharding. There's a security tradeoff with sharding, as there is with almost any scalability solution - it's probably best to wait until our network topology is strong enough to handle something like sharding before implementing it. If yu don't have a lot of nodes, sharding can reduce the security of a network significantly'

Asher Manning, [30.12.17 13:07]

'If we were to implement sharding it would most likely be identical to Zilliqa's mechanism, at the lowest level. Zilliqa allows non-overlapping transactions to be split amongst different nodes, which use a proportionally reduced PoW scheme. We had the same idea floating around in around september 2016, but as Nathan just mentioned, sharding isn't the only scaling solution. It also isn't appropriate until a very strong network topology is achieved. Not practical for a newer blockchain, it needs a lot of nodes to be secure'

Asher Manning, [30.12.17 13:07]

'Ethereum's state sharding is not relevant to us, we do not have shared state so we do not need a solution to splitting and merging shared state.'

Asher Manning, [30.12.17 13:07]

'There are some interesting things that can be done with sharding + multi-hash mining, for example splitting transactions across PoW algorithms instead of using reduced difficulties'

Asher Manning, [30.12.17 13:07]

'In general though we don't like to throw buzzwords around and commit to doing things that aren't actually helpful. The scalability meaures that we've commited to (parallel validation, statelessness, merkelised contracts, ACS) are all appropriate to our architecture and do not impose security risks. Other ideas that float around like lightening networks, sharding, etc. are major influences on network topology, and can seriously impact the degree of decentralisation and security if they are not done correctly. We are thinking about privacy as well, and so we are very cautious regarding adding in new network features that might prevent us from doing other things later down the road.

 Scalability is not about having the best buzzword collection, it's about having good throughput, good security, and good decentralisation. Sometimes it's about having good privacy too. Acheiving this requires a lot of research and thought to make sure that it will all work as planned, it's not as simple as just picking and mixing network features.'

Asher Manning, [30.12.17 13:07]

'Linear scaling is a desirable property, but keep in mind that the 'efficiency' gain ( I assume you mean throughput gain?) from sharding is at odds with security - as you have more shards in a network, the hashpower required by an attacker is reduced. 

Shared state means that contracts can access a shared store of data and interact arbitrarily - in Ethereum, this shared state is the EVM. It is difficult for ethereum to shard because contracts change the state of the EVM, so each shard has a different state when the shards need to be merged, and merging these shards is a complex problem. 

If you do not have a shared state, then you only need to merge UTXO sets, which is not so difficult.

The network topology refers to how many nodes there are in the network and how much hashpower the network has. Since sharding reduces the amount of hashpower needed by an attacker, it is practical only if a network has a high amount of hashpower, distributed across several nodes to shard across.'