[DISCUSS] Providing smaller compressed source tarballs

[Chí-Thanh Christopher Nguyễn <ch...@gentoo.org>]

Hi,

Currently, the 3.4.0 source tar.bz2 is around 240 MiB. Providing an xz
compressed tarball instead would reduce download size for users who
build AOO from source.

Informal comparison between different compressors:
1697468416 aoo-3.4.0-incubating-src.tar
 251417600 aoo-3.4.0-incubating-src.tar.bz2
 207922176 aoo-3.4.0-incubating-src.tar.xz-6e
 199165952 aoo-3.4.0-incubating-src.tar.xz-7e
 190512128 aoo-3.4.0-incubating-src.tar.xz-8e
 187145216 aoo-3.4.0-incubating-src.tar.xz-9e
 168845312 aoo-3.4.0-incubating-src.tar.zpaq-m3-b256
 168816640 aoo-3.4.0-incubating-src.tar.zpaq-m3
 159088640 aoo-3.4.0-incubating-src.tar.zpaq-m4-b256

ZPAQ compresses even better, but compression/decompression is very slow.
It seems that xz -8e would be a good compromise between CPU/memory usage
and compression ratio. It results in a 24% smaller size compared to
bzip2 -9.

Do you agree or have suggestions/concerns?


Best regards,
Chí-Thanh Christopher Nguyễn

Re: [DISCUSS] Providing smaller compressed source tarballs

[Chí-Thanh Christopher Nguyễn <ch...@gentoo.org>]

Andrea Pescetti schrieb:
> Indeed, download size for the sources is not a big problem (size of
> binaries is the issue) but thanks for the analysis: it will be useful if
> someone wishes to redistribute OpenOffice on CD-ROM and include source
> code, which as far as I know is optional with the current license.

Thanks for your reply.

For users of source based Linux distros like Gentoo, or users that
choose to build from source on their otherwise binary distro, the source
tarball size is more important. Even if the number of such users is
small, they can be made happier at almost no cost. The bz2 tarball could
even be dropped as it compresses worse and takes longer to decompress
than xz, keeping gz for those who value decompression speed over
download size.


Best regards,
Chí-Thanh Christopher Nguyễn

Re: [DISCUSS] Providing smaller compressed source tarballs

[Andrea Pescetti <pe...@apache.org>]

On 10/05/2012 Chí-Thanh Christopher Nguyễn wrote:
> Currently, the 3.4.0 source tar.bz2 is around 240 MiB. Providing an xz
> compressed tarball instead would reduce download size for users who
> build AOO from source.

I remember making a similar assessment years ago for a variant of the 
OOo 2.x sources. The license required that we distributed the source 
code in CD-ROMs (or made it available in other ways, but this was the 
more straightforward way to comply) and in the end we settled for 7-ZIP.

Indeed, download size for the sources is not a big problem (size of 
binaries is the issue) but thanks for the analysis: it will be useful if 
someone wishes to redistribute OpenOffice on CD-ROM and include source 
code, which as far as I know is optional with the current license.

Regards,
   Andrea.