How you can have real-time database modifications if transactions require acceptance by all others

Question

I'm trying to understand the current implementation of smart contracts and transactions and am wondering about real-time behavior. I read that it can take up to 10 minutes for a transaction to become part of a block, and read here that "you have to create a so-called transaction which has to be accepted by all others". So I don't see how you could have real-time transactions and built say a Todo list application where each time you created a todo it would instantly save to the database and upon refresh would be there, and it would charge the user per todo (let's say) immediately. Wondering if this is possible, or if it's been considered, or why it's not possible, or how you would handle a situation like this.

Answer 1

I'm unsure where you have found out that it would take 10 minutes for a transaction to be in a block. It does not take 10 minutes. Currently the average time is a bit less than 20 seconds (https://ethstats.net/).

According to the network everything is ready once a transaction is mined in a block but the consensus system dictates that a block is actually never final. In theory even the first block of the network might get reverted due to consensus chain reorganization. In practice though a block is final after a few other blocks are mined on top of it.

It's basically up to you to decide when do you accept for a block to be final enough with a big enough certainty. Some exchanges for example require 10 blocks (10 confirmations) on top of the block to consider it final enough to be accepted as a non-revertible transaction.

So, in theory the blockchain is some "20 seconds away from real-time". 20 seconds might not be too long an interval for a todo application. But you always have to remember that there is the consensus tradeoff to be made: how many blocks do you want to wait before you trust the block to be final enough.

REMEDIES

There are ways to circumvent this "slow" transaction time. For example:

1) Do some of the processing outside blockchain. Obviously all such processing is much faster. Then just update data in the blockchain at certain intervals or so.

2) Side chains / state channels. These are rather new approaches but they partially utilize the first method; some of the processing is done in for example a side chain which performs much faster. These solutions are mostly valid only for transactions between known participants (so a lot of transactions back and forth between known participants) and not for "random" transactions to random participants.