Hey Daniel, Avalanche certainly has made some neat steps recently. From a

Question

Hey Daniel, Avalanche certainly has made some neat steps recently. From a

tech point of view, I have a few issues with their choices.

- they started with a very low latency - they say they have sub-second finality.

Sub-second finality is optimistic, at best. There's no guarantee a transaction is final, unless a geographically dispersed network of nodes has time to receive it, study it, reach consensus on it, inform the network that it is indeed final.

Outside of some magic on behalf of Avalanche, those sub-second confirmation times look like they can be a lot of trouble, from simply missed rounds where blocks won't get minted, to more serious trouble such as double spending, interrupting the network, etc. They'll probably learn that the hard way.

At Elrond, we tested things in a public & highly dispersed geographical environment, with random computers we knew nothing about (as is of course the case in decentralized networks) - we saw that "lazy" computers could cause disruption, slow-downs, etc. I'm sure the Avalanche team would notice the same, if they tested that way.

We decided to be 100% focused on security and stability, so we debuted with a 6s latency. We could probably decrease to 5, maybe even 4? The lower you get, the more risk you take. We will of course let the network of validators decide. But I think real adoption will only focus on low risk networks.

- the Avalanche consensus involves selecting the nodes with the highest stake as preferred validators. That's not as bad as delegated proof of stake, but it's not far from it. In Elrond, all nodes are equal.

- The Avalanche architecture stitches 4 different chains together. Security chain, transactions chain, private chains chain and smart contracts chain. They also have different consensus protocols 😵

Elrond has multiple "chains" as well, but they're all identical - shards - they are more because each one can handle a certain amount of load. More throughput needed: add more shards. That's called scalability: we atm handle 15,000 TPS with 3 shards, can scale to a lot more (we did 263k TPS with 50 shards on the public testnet).

Avalanche can't do that. They have 4500 TPS on their transaction chain. Curious to see what their smart contracts chain is capable of.

Also I think developers have the extra hurdle of managing those 4 individual chains for the specific tasks they are trying to accomplish.

TL;DR - meh.