[Debian-l10n-devel] on the size of ftp..dists..i18n

[Peter Palfrader]

Hi,

I am playing with building a snapshot service similar to
snapshot.debian.net and I used this to look at what stuff is making the
mirror push so large.

It appears that one of worst offending directories in all the ftp
archive is the dists/sid/main/i18n directory, closely followed by the
lenny/main/i18n directory.  (Only being beaten by the directory that
holds the Contents files.)

They add up to a 70 megs change every time they get updated, which right
now appears to be twice a day, resulting in 140 megs (a significant
amount of each mirror run).

Looking at that directory I see you ship every single file three times.
Once uncompressed, once gzip compressed and once bzip2 compressed.
| -rw-r--r-- 2 dak debadmin 4868419 Feb  3 12:40 Translation-de
| -rw-r--r-- 2 dak debadmin 1322033 Feb  3 13:52 Translation-de.bz2
| -rw-r--r-- 2 dak debadmin 1853533 Feb  3 13:52 Translation-de.gz

If you could get rid of the gzip and uncompressed version (and if the
above example is representative) you could cut down the change by
over 80% to a meager 16% of what you use now.

Is that an option?

weasel
-- 
                           |  .''`.  ** Debian GNU/Linux **
      Peter Palfrader      | : :' :      The  universal
 http://www.palfrader.org/ | `. `'      Operating System
                           |   `-    http://www.debian.org/