Gavin Andresen: Designed for Success

I just listened to a Bitcoin podcast with Emin Gün Sirer and Ittay Eyal from Cornell University.

They are doing great work; a full simulation of the Bitcoin network is a really fantastic idea, and I plan to do a lot of testing and optimization with the tools they develop. I also plan to write about their Bitcoin NG ideas... but not right now.

Listening to this podcast and hearing complaints from one of the core committers made me realize there was a fundamental disagreement about the design of the protocol.

The most successful protocols are forward-looking. When the IP protocol was invented in the 1970s, the idea of ​​4 billion computers connected to a network was ridiculous. But the designers were forward-looking and used 32-bit IP addresses. The protocol grew from a small research project to today's global network. 40 years later, the total number of available IP addresses is facing exhaustion.

I applaud Gün and Ittay for their scientific approach to the Bitcoin network, and for establishing metrics to evaluate the performance of work or proposals. But I think it is too easy to be stuck with the current implementation of the Bitcoin network, and I do not think that the current implementation in Bitcoin Core should dictate higher-level protocol design.

I believe that the design of the agreement should be forward-looking and should not rely solely on individual performance work.

I understand the desire to be conservative and test within the limits of any protocol. The most common criticism I hear from some people about BIP101 is, “You haven’t tested a network with gigabyte-sized blocks,” and I wonder if the IP designers have also been complained about, “We haven’t tested a network with a billion computers connected,” and if those IP designers hadn’t used such a forward-thinking protocol, I don’t know what Internet Protocol we would have today.

I keep hearing that bigger blocks may drive up the mining centralization problem, but I wrote about this earlier this year and still don’t see a convincing argument unless you believe that the current P2P protocol is set in stone and won’t change in the future.

I’m going to work on protocols that make transactions and block propagation better across the network, because a better protocol is necessary if miners are going to want to create larger blocks (we already have one in the form of Matt Corallo’s “fast relay network”, which is a big reason why most mining pools are willing to create 1MB blocks) but I think it would be a mistake to wait for three reasons:

1. First, for any protocol change, it typically takes about six months to be deployed across the entire network.

2. Second, because someone else might have a better idea than me. With a 1MB block size limit, there is little incentive to optimize transactions and block propagation (why spend so much time writing code that can be efficiently implemented with a larger block size?)

3. Finally, miners are not stupid, and when the slush pool produced a 900+KB block that forked the chain, the largest miners immediately agreed to produce small blocks until the Bitcoin Core software was fixed.

－－－－

Original article: http://gavinandresen.ninja/designing-for-success
By GAVIN ANDRESEN
Compiled by: Satuoxi
Editor: Satuoxi
Source (translation): Babbitt Information