Date: Mon, 24 Feb 2014 13:29:31 -0500
From: Jim Paris <jim@jtan.com>
To: Mark Friedenbach <mark@monetize.io>
Message-ID: <20140224182931.GA16588@psychosis.jim.sh>
References: <5306BCC1.8040004@monetize.io>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <5306BCC1.8040004@monetize.io>
Sender: jim@jim.sh
User-Agent: Mutt/1.5.20 (2009-06-14)
Cc: Bitcoin Dev <bitcoin-development@lists.sourceforge.net>
Subject: Re: [Bitcoin-development] Base-32 error correction coding
Precedence: list

Mark Friedenbach wrote:
> What follows is a proposed BIP for human-friendly base-32
> serialization with error correction encoding.
...
> 2. Automatic correction of up to 1 transcription error per 31 coded
> digits (130 bits of payload data). For a 256-bit hash or secret key,
> this enables seamless recovery from up to two transcription errors so
> long as they occur in separate halves of the coded representation.

Can we do better than correcting single transcription errors?  I'd
imagine that transposition of two adjacent characters, or insertion or
deletion of a single character, would be very common.  At the very
least, a transposition could be corrected by interleaving the two
"halves of the coded representation", e.g.:

   ABABABABABABABABABABABABABABABABABABABABABABABABABABABABABABAB

insead of
   
   AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABBBBBBBBBBBBBBBBBBBBBBBBBBBBBBB

Jim