On 12/9/05, <b class="gmail_sendername">Mikael Eriksson</b> <<a href="mailto:mikael_miffe_eriksson@yahoo.se">mikael_miffe_eriksson@yahoo.se</a>> wrote:<div><span class="gmail_quote"></span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
On Thu, Dec 08, 2005 at 06:18:00PM -0800, Matt Zimmerman wrote:<br>> On Wed, Dec 07, 2005 at 04:58:28PM +0200, Timo Jyrinki wrote:<br>> > Speaking of language files, those also compress better with bz2 even though they
<br>> > are in mo format:<br>> > -rw-r--r-- 1 root root 1356095 2005-12-07 16:51 LC_MESSAGES-fi.tar.bz2<br>> > -rw-r--r-- 1 root root 1590691 2005-12-07 16:52 LC_MESSAGES-fi.tar.gz<br>> ><br>> > (compressed etch's /usr/share/locale/fi/LC_MESSAGES to get those)
<br>><br>> That's why the language pack .debs are already compressed with bzip2.<br>><br>A repack of language-pack-gnome-* shows that most would be smallest when<br>compressed with 7z. And the rest is smallest with gzip -9.
<br>See the attached file.<br></blockquote></div><br>
An idea: the compression tests could be done automatically by a
'repacking' script on the central repository. It's a computationally
intensive task, but it could be done in the background. The script
would do the following:<br>
<br>
-- get the original .deb<br>
-- repack the components with a number of different tools<br>
-- select the smallest ones<br>
-- generate a new deb file<br>
<br>
Although slow, this is at least an automatic procedure -- something you
can throw at a server to work over 24x7 day after day, perhaps using
otherwise unused cycles. The savings in bandwidth (both server side and
user side) could potentially pay for it. And we would avoid putting one
more strain on the back of packagers (testing to check which tool
compresses better) -- they could always work with .tar.gz for
development.<br>
<br>
(After writing I realized that the repacker script can also be
distributed to do this automatically... but for some reason I believe
this is something that can be handled by a server in the repository.)<br>
<br>-- <br>Carlos Ribeiro<br>Consultoria em Projetos<br>blog: <a href="http://rascunhosrotos.blogspot.com">http://rascunhosrotos.blogspot.com</a><br>blog: <a href="http://pythonnotes.blogspot.com">http://pythonnotes.blogspot.com
</a><br>mail: <a href="mailto:carribeiro@gmail.com">carribeiro@gmail.com</a><br>mail: <a href="mailto:carribeiro@yahoo.com">carribeiro@yahoo.com</a><br><br>