From: "Luke-Jr" <luke@dashjr.org>
To: Mark Friedenbach <mark@monetize.io>
Date: Wed, 13 Mar 2013 18:04:08 +0000
User-Agent: KMail/1.13.7 (Linux/3.7.3-gentoo; KDE/4.9.5; x86_64; ; )
References: <201303131256.30144.luke@dashjr.org>
	<CACh7GpE09hqCPL6rtdC6EZzohM5QHb+0SdFoD8MzPSqf7=6hOA@mail.gmail.com>
In-Reply-To: <CACh7GpE09hqCPL6rtdC6EZzohM5QHb+0SdFoD8MzPSqf7=6hOA@mail.gmail.com>
MIME-Version: 1.0
Message-Id: <201303131804.09339.luke@dashjr.org>
Content-Type: Text/Plain;
  charset="iso-8859-15"
Content-Transfer-Encoding: 7bit
Cc: "bitcoin-development@lists.sourceforge.net"
	<bitcoin-development@lists.sourceforge.net>
Subject: Re: [Bitcoin-development] 0.8.1 ideas
Precedence: list

On Wednesday, March 13, 2013 5:41:29 PM you wrote:
> I'm not sure I understand the need for hard forks. We can get through this
> crisis by mining pool collusion to prevent forking blocks until there is
> widespread adoption of patched clients.

Anything requiring widespread adoption of patched clients *is by definition* a 
hard fork.

> Proposal:
> 
> 1) Patch the pre-0.8 branches to support an increased lock count, whatever
> number is required to make sure that this problem never shows up again at
> the current block size (I defer to Luke-Jr and gmaxwell's numbers on this).

This is a hard fork.

The only way to avoid a hard fork is to apply the existing lock limit to all 
clients forever. That would be fine, except that pre-0.8 clients cannot reorg 
N blocks without dividing that limit by (N * 2) + 1; that leaves us with the 
limit of around 1000 locks per block on average. Each transaction uses at 
least 3 locks on average (many times more). So about 300 transactions per 
block. This is a much smaller limit than the 1 MB we've been assuming is the 
bottleneck so far, and the need to increase it is much more urgent - as Pieter 
noted on IRC, we are probably already using more than that even ignoring DP 
spam. The only reason pre-0.8 clients have survived as well as they have thus 
far is because the blockchain has managed to avoid very deep reorgs.

Luke