[arch-dev-public] [pacman-dev] GZip compression ratios

Dan McGee dpmcgee at gmail.com
Fri Feb 8 00:47:45 EST 2008


Oops, left the dev list off my original reply. Text is below.

On Feb 7, 2008 11:24 PM, Travis Willard <travis at archlinux.org> wrote:
> Just doing an update, and I noticed this:
>
> Targets: coreutils-6.10-1  bash-3.2.033-2  cairo-1.4.14-1
> gtk2-2.12.7-1  gcc-libs-4.2.3-3  mpfr-2.3.1-1  gcc-4.2.3-1
> mesa-7.0.3rc1-1  acroread-8.1.2-1  automake-1.10.1-2
>          kernel26-2.6.24-2  catalyst-8.01-3  curl-7.18.0-1
> devtools-0.5-1  dhcpcd-3.2.1-1  e2fsprogs-1.40.5-1  faac-1.26-1
> filesystem-2007.11-6  fakeroot-1.9.2-1
>          findutils-4.2.32-1  flex-2.5.33-4  librsvg-2.20.0-1
> libxmu-1.0.4-1  gimp-2.4.4-1  git-1.5.4-1  gpgme-1.1.6-1
> groff-1.19.2-4  kde-common-3.5.8-3  lm_sensors-3.0.1-1
>          kdebase-3.5.8-3  lftp-3.6.3-1  libgnomecups-0.2.3-1
> libxml2-2.6.31-1  libgnomeprint-2.18.3-1  libgnomeprintui-2.18.2-1
> libidl2-0.8.10-1  libmysqlclient-5.0.51-2
>          libtasn1-1.3-1  libtool-1.5.26-1  lua-5.1.3-1  lvm2-2.02.33-1
>  man-1.6f-2  man-pages-2.77-1  mysql-clients-5.0.51-3  nasm-2.01-1
> orbit2-2.14.12-1  pcre-7.6-2
>          php-5.2.5-3  pil-1.1.6-4  postgresql-libs-8.3.0-1
> sdl_mixer-1.2.8-2  shared-mime-info-0.23-1  snownews-1.5.8-1
> spamassassin-3.2.4-2  sqlite3-3.5.5-1  sudo-1.6.9p12-1
>          tk-8.5.0-2  wget-1.11-1  which-2.19-2  xfsprogs-2.9.5-1
> xine-lib-1.1.10-1  xorg-server-1.4.0.90-6
>
> Total Download Size:    194.03 MB
> Total Installed Size:   194.39 MB
>
> Does gzip compression really give us so little as to only save 0.3 MB
> over uncompressed for so many packages?  I figured the difference
> would be a bit more - does it suck for compressing binary data?
> Should we consider moving to a different, more compressed archive
> format?  Or am I just making a mountain out of a molehill here?

No, its a bunch of bullshit.

If/when the repo db scripts get rewritten, we would actually have
installed sizes available in our PRIMARY repos. Currently the only
repo that has them is community, so what you see above is that the
installed size of your community packages just happens to be slightly
greater than the download size of all of the packages, so installed
size is shown. It is a terrible heuristic, see this commit queued for
3.2 for details:
http://projects.archlinux.org/git/?p=pacman.git;a=commitdiff;h=93a3050ed9e9735352013014da52adcdf896f95f

Thus, anything in core or extra will lie to you.

-Dan




More information about the arch-dev-public mailing list