[rdiff-backup-users] rdiff-backup gzip compression

[Niall O Broin]

I'm curious as to what exactly is doing to compress metadata files. I recently 
had occasion to unzip a couple of them (I wanted to remove a file - yes, I know 
it's dodgy, but I had backed up a 30+ GB file I didn't want to keep) and when I 
compressed them again, some were larger than the originals. Not by much of 
course, and in the context of an rdiff backup the disk space used isn't the 
issue at all - I'm just curious as to what rdiff-backup is doing that the gzip 
standalone program isn't:

address@hidden:/tmp/mirror_metadata# ls -l
total 9828
-rw-r--r-- 1 root root 9554194 Nov  1 12:01 
mirror_metadata.2013-10-23T01:54:06+02:00.snapshot.gz
-rw------- 1 root root  486893 Nov  1 12:01 
mirror_metadata.2013-10-31T03:06:42+01:00.diff.gz
address@hidden:/tmp/mirror_metadata# gunzip *
address@hidden:/tmp/mirror_metadata# ls -l
total 70800
-rw-r--r-- 1 root root 69103098 Nov  1 12:01 
mirror_metadata.2013-10-23T01:54:06+02:00.snapshot
-rw------- 1 root root  3316456 Nov  1 12:01 
mirror_metadata.2013-10-31T03:06:42+01:00.diff
address@hidden:/tmp/mirror_metadata# gzip --best *
address@hidden:/tmp/mirror_metadata# ls -l
total 9832
-rw-r--r-- 1 root root 9554194 Nov  1 12:01 
mirror_metadata.2013-10-23T01:54:06+02:00.snapshot.gz
-rw------- 1 root root  488606 Nov  1 12:01 
mirror_metadata.2013-10-31T03:06:42+01:00.diff.gz

The snapshot file ended up the exact same size, but the diff file is a little 
(about 0.35% ;-) ) smaller. I had a read of 
http://www.gzip.org/zlib/rfc-gzip.html and if anything the original file should 
have been marginally bigger, as it, in contravention of the file format 
specification, stores the full path of the compressed file in the filename 
component of the header, thus needing a few extra bytes.

So, any ideas what might be happening?



Kindest regards,



Niall  O Broin