There is no 51% attack in PoS system. Rather there is a more than 1/3rd attacker problem.

What folks?

Not true

How is that not true? In grin you only have to download the block headers and only the last 5000 blocks (called horizon) for a full sync. Bandwidth saving is even greater than 99%.

Wow wow wow

Yes people can duck with a network by cutting cables. This works with all internet protocols.

If 99% of Bitcoin's hashrate gets taken off the internet, it keeps going without any problem. That's the beauty of a decentralized cryptocurrency. There's no central node and it's supremely fault tolerant.

Algorand has a lot of interesting technology and may have some use cases. But it's more like a distributed database / voting system than a true cryptocurrency.

That is not true. If network partitions happen in bitcoin you are utterly ducked. The system will fork and good luck to recover from that with all the double spendings.

It is already dying

Algorand

If no fees, what are the incentives to accept transactions in a block?

Also, what stops you from spamming the network?

A small PoW on each block, however for some reason it can be precomputed which allows you to precompute a spam attack (which has happened from the CEO of BrainBlocks recently..)

It's an interesting issue, although it's a good question if it is possible to stop network spam. Bitcoin's solution to network spam is to limit the amount of transactions and charge a fee, with the fee going up the higher the demand for a transaction.

I don't see this setup working with Nano, as much of the protocol is designed around not having fees. After all, deciding who gets the fees and splitting it (among top representatives?) causes even more processing on the network.

I've had concerns in this area as well, as far as I know the team performed some stress tests and the network can do 100+ transactions per second, maybe more. There is also a small computation that the hardware needs to perform before sending the transaction, which is only a small guard against spam but pretty useful.

afair peak of stress test on main network was >300TPS, on test network is was >1500TPS. there exists dedicated software to create wallets and transaction using GPU, just for stress testing.

The stake in the network. Each transaction is validated by representatives whose votes are proportional to their stake in the network. An attacker would need to gain 50% of the network value to be able to do some damage.

You are not answering the question.

Users with high stakes generally want the network to run so their stakes aren't worthless. At least, that's the theory behind PoS.

PoS by itself doesn't incentivize you to include transactions. That's called the "nothing at stake" problem.

When you look at the big cryptographers behind vault and algorand you understand why.

No it doesn't. Coda can work with Algorand's byzantine agreement protocol.

Sure, but Coda targets a different scalability bottleneck.

Hence why it can work with algorand.