doublespend – Minimal requirements for the proof of work

First, to address a few misconceptions:

• The nonce is part of the block, together with all its transactions (including the miner payout address), timestamp, previous block hash, … Changing any of these will completely change the block hash.

• The difficulty is not the number of 0 bits required in the hash. Rather, the entire block hash is interpreted as a 256-bit number, and compared to something called the “target”, also a number. If the hash is lower than the target, the proof of work is valid. Difficulty adjustment should really be called target adjustment: every 2016 blocks, the target is increased or decreased to keep the average rate of blocks at 1 per 10 minutes. Difficulty is the ratio of the highest possible (=easiest) target divided by the actual target.

• The accepted chain is not necessarily the longest chain; rather, it is the chain with the most accumulated work. Work here is (roughly) defined as the sum of the difficulties of the blocks in the entire chain. This means that the rate of producing blocks is immaterial for deciding who creates the winning chain, as faster blocks also have lower difficulty.

Then as for your real question: what are the harmful effects of a shorter inter-block time?

The purpose for the time between blocks is giving the network time to propagate the newly-found block, so that all nodes and miners are on the same page.

Imagine it takes 1 second for a block to propagate through the network, and for miners to start building on top of it (so this number includes network latency, but also validation time, the time it takes to build a new candidate set of transactions to include in a next block, and to distribute this information to hashers).

Miner A finds a block, and starts propagating it, so miners B, C and D all learn about this block 1 second later. During that 1 second, B, C, and D are effectively wasting their time, as they are trying to build on top of the previous block still. However, A does not waste any time: it can switch to working on top of the new block immediately. This creates a tiny advantage for A, proportional to the probability that another block is found within this 1 second propagation time.

With an inter-block time of on average 10 minutes, this advantage is tiny (roughtly 0.17% chance to find a block in 1 second), but it grows very significantly as the inter-block time and propagation time get closer to each other:

• 600s vs 1s: 0.17%
• 60s vs 1s: 1.7%
• 10s vs 1s: 9.5%
• 1s vs 1s: 63%

This has a number of effects:

• Even without malice, shorter block times proportionally benefit larger miners more. This is counter to the intent of fair mining, where anyone can enter the mining space and get benefits proportional to their investments.
• It can be intentionally exploited more easily through so-called “selfish mining” attacks, which are always possible, but propagation times worsen it.
• A prospective 51% attacker (one who ignores blocks produced by honest miners and only builds on top of their own) ignores this propagation cost. So the probability for a block to be found while the previous block is still propagating directly translates to a benefit 51% attackers have.