The context is that I am looking into a business plan to implement ERC20-token based payment at standard merchant terminals, plus corollary services including loyalty, and looking at transaction throughput.
Apparently VISA can process tens of thousands of transactions per second whereas Ethereum is limited to the tens, as long as Proof-of-Work is being used, and assuming we are talking about the core network only not Raiden etc. (Would love to hear challenges to this premise).
Looking at payment channels as a way of increasing throughput, I am trying to understand how in real life scenarios these help much.
So for example: Day 1: Cardholder receives a physical card associated with an ERC20 balance of say $100 (1:1 tokens for example). This amount is held in escrow for the payment channel. For the sake of this example, the payment channel is with a single merchant, not an aggregating provider or merchant network.
Day 2: Cardholder goes to merchant and buys some shopping. This event is not likely to happen at a rate higher than say once per day on average. The payment channel is updated offchain with signed messages (details reserved)
Day 3: Cardholder repeats, shopping...
Day N: The card balance is used or the cardholder wishes to top up the balance, closing the payment channel. The Ethereum blockchain is updated with the net balance.
So, in this kind of scenario, in order to get only 20 or 30 times the throughput, the payment channel has to be open for one month. There would be a number of challenges with how to batch the payment channel closures, as a merchant (eg: Walmart) with say one million cardholders could result in delays closing and reopening channels.
Browsing the net, I may have seen claims of greatly increased throughput via Raiden/Lightning network for point of sale processing, but I am at a bit of a loss as to how this could work in practice. The way I see it, is that a payment channel, like a side-chain, is in general a way of batching transactions so that the Ethereum blockchain can a) act as a settlement engine b) process a smaller number of transactions.