Date: Fri, 31 Jul 2015 08:39:43 +0000
Message-ID: <20150731083943.Horde.68uT9J78H_PdIgIwQP5frA1@server47.web-hosting.com>
From: jl2012@xbt.hk
To: "bitcoin-dev@lists.linuxfoundation.org"
	<bitcoin-dev@lists.linuxfoundation.org>
User-Agent: Internet Messaging Program (IMP) H5 (6.1.4)
Content-Type: text/plain; charset=UTF-8; format=flowed; DelSp=Yes
MIME-Version: 1.0
Content-Disposition: inline
Subject: [bitcoin-dev] A compromise between BIP101 and Pieter's proposal
Precedence: list

There is a summary of the proposals in my previous mail at  
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2015-July/009808.html

I think there could be a compromise between Gavin's BIP101 and  
Pieter's proposal (called "BIP103" here). Below I'm trying to play  
with the parameters, which reasons:

1. Initiation: BIP34 style voting, with support of 750 out of the last  
1000 blocks. The "hardfork bit" mechanism might be used:  
http://lists.linuxfoundation.org/pipermail/bitcoin-dev/2015-July/009576.html

Rationale: This follows BIP101, to make sure the new chain is secure.  
Also, no miner would like to be the first one to mine a large block if  
they don't know how many others would accept it.

2. Starting date: 30 days after 75% miner support, but not before  
2016-01-12 00:00 UTC

Rationale: A 30-day grace period is given to make sure everyone has  
enough time to follow. This is a compromise between 14 day in BIP101  
and 1 year in BIP103. I tend to agree with BIP101. Even 1 year is  
given, people will just do it on the 364th day if they opt to  
procrastinate.

2016-01-12 00:00 UTC is Monday evening in US and Tuesday morning in  
China. Most pool operators and devs should be back from new year  
holiday and not sleeping. (If the initiation is delayed, we may  
require that it must be UTC Tuesday midnight)

3. The block size at 2016-01-12 will be 1,414,213 bytes, and  
multiplied by 1.414213 by every 2^23 seconds (97 days) until exactly  
8MB is reached on 2017-05-11.

Rationale: Instead of jumping to 8MB, I suggest to increase it  
gradually to 8MB in 16 months. 8MB should not be particularly painful  
to run even with current equipment (you may see my earlier post on  
bitctointalk: https://bitcointalk.org/index.php?topic=1054482.0). 8MB  
is also agreed by Chinese miners, who control >60% of the network.

4. After 8MB is reached, the block size will be increased by 6.714%  
every 97 days, which is equivalent to exactly octuple (8x) every 8.5  
years, or double every 2.9 years, or +27.67% per year. Stop growth at  
4096MB on 2042-11-17.

Rationale: This is a compromise between 17.7% p.a. of BIP103 and 41.4%  
p.a. of BIP101. This will take us almost 8 years from now just to go  
back to the original 32MB size (4 years for BIP101 and 22 years for  
BIP103)

SSD price is expected to drop by >50%/year in the coming years. In  
2020, we will only need to pay 2% of current price for SSD. 98% price  
reduction is enough for 40 years of 27.67% growth.
Source: http://wikibon.org/wiki/v/Evolution_of_All-Flash_Array_Architectures

Global bandwidth is expected to grow by 37%/year until 2021 so 27.67%  
should be safe at least for the coming 10 years.
Source:  
https://www.telegeography.com/research-services/global-bandwidth-forecast-service/

The final cap is a compromise between 8192MB@2036 of BIP101 and  
2048MB@2063 of BIP103


-----------------------------------

Generally speaking, I think we need to have a faster growth in the  
beginning, just to normalize the block size to a more reasonable one.  
After all, the 1MB cap was introduced when Bitcoin was practically  
worthless and with inefficient design. We need to decide a new  
"optimal" size based on current adoption and technology.

About "fee market": I do agree we need a fee market, but the fee  
pressure must not be too high at this moment when block reward is  
still miner's main income source. We already have a fee market: miners  
will avoid building big blocks with low fee because that will increase  
the orphan risk for nothing.

About "secondary layer": I respect everyone building secondary layer  
over the blockchain. However, while the SWIFT settlement network is  
processing 300tps, Bitcoin's current 7tps is just nothing more than an  
experiment. If the underlying settlement system does not have enough  
capacity, any secondary layer built on it will also fall apart. For  
example, people may DoS attack a Lightening network by provoking a  
huge amount of settlement request which some may not be confirmed on  
time. Ultimately, this will increase the risk of running a LN service  
and increase the tx fee inside LN. After all, the value of secondary  
layer primarily comes from instant confirmation, not scarcity of the  
block space.