From: Neal Kreitzinger <nkreitzinger@gmail.com>
Date: Wed, 25 Apr 2012 21:15:32 GMT
Subject: how to analyze binary delta compression
Message-ID: <jn9phl$t7s$1@dough.gmane.org>

I need to assess the delta compression efficiency of various types of 
binaries.  Can I do it this way:

(1) Run git diff between two versions of a binary and pipe the result to 
a file.  (Sample different types of binaries and changes.)
(2) Assess the output by size or format or content to determine how well 
it delta differenced?  How would I determine this?  What are good 
absolute delta sizes?  What are good delta size ratios compared to total 
file size?  Is the number of hunks in the delta important?
(3) Compress the delta file with gzip (or something) and then assess the 
compression ratio?  What is a good compression ratio?  What are good 
delta compressed absolute sizes?  What are good delta compressed 
relative sizes to total file size?

I'm sure I may be missing some things here.  I'm new to assessing delta 
compression efficiency of binaries in general, and specifically for git. 
  Please advise.

v/r,
neal