[tex-live] Re: Re: Re: [tug-board] Any news of DVDs ?

[Giuseppe Bilotta]

On Wed, 1 Feb 2006 00:58:23 +0100, Reinhard Kotucha wrote:

>>>>>> "Karl" == Karl Berry <karl at freefriends.org> writes:
> 
>   >     Incidentally, 7-zip even has its own format (7z) which in many
>   > cases packs much better than anything else.
> 
>   > Yeah.  If someone would like to try 7zipping the TL iso, I'd be
>   > interested to know how it fares, both in how well it compresses
>   > and in decompression time.
> 
> sizes
> -----
> The numbers in braces denote compression gain.  
> 
>     100 * (1-(compressed_size/iso_size))
> 
> 714938368  TeXLive-2005-install.iso
> 684287622  TeXLive-2005-install.iso.bz2 (4.2872%)
> 682955930  TeXLive-2005-install.iso.zip	(4.4735%)
> 682955793  TeXLive-2005-install.iso.gz  (4.4735%)
> 677460837  TeXLive-2005-install.iso.7z  (5.2421%)
> 
> I suppose that in Hans' example the content of the ISO-images had not
> been compressed.  Hence, compression gain depends much more on the
> program than it does when the content of the image is already
> compressed.

Yes, re-compression of an already compressed source isn't going to
help much. The ideal would be if it was possible to compress the data
*in* the ISO, rather than the ISO itself, with 7z.

Now, the interesting part about 7z is that it has gathered a lot of
interest, so although it was initially only available on Windows, it's
being ported to POSIX systems. NSIS, the former NullSoft Installer,
uses LZMA since version~2.

It's probably too late to do anything about it for this ISO, but for
the future it may be a good thing to evaluate: IMO it makes sense to
use it as packing format for multiplatform distributions, so it may be
useful to cram more stuff on the images (rather than for packing the
ISO image itself).

And yes, this would have to be brought up with the CTAN maintainers
too. Or maybe we should wait for 7z to get a little more widespread
...

> de-compression
> --------------
> ~ $ time bunzip2 TeXLive-2005-install.iso.bz2 
> real    2m25.975s
> user    2m2.304s
> sys     0m4.644s
> 
> ~ $ time 7z -y e TeXLive-2005-install.iso.7z
> real    1m33.379s
> user    1m7.852s
> sys     0m4.216s
> 
> ~ $ time unzip -o TeXLive-2005-install.iso.zip 
> real    0m30.398s
> user    0m6.184s
> sys     0m3.260s
> 
> ~ $ time gunzip TeXLive-2005-install.iso.gz 
> real    0m30.952s
> user    0m6.492s
> sys     0m3.724s
> 
> I must admit that I did not read all the documentation.  I don't use
> zip very often and installed 7-zip just after I read your mail.  The
> 7z manual page is quite short and refers to additional documentation.
> At least there is a command line option which allows you to select a
> compression algorithm.  Maybe some of the programs can be tweaked to
> fit a particular need.

Yes, the 7z format can make use of different compression algorithms,
but the most commonly used is the general purpose LZMA, which uses the
same concepts as zip and gzip (i.e. Lempel-Ziff) but allows a much
wider range of tuning options to improve the compression ratio
(therefore getting nearer the 'perfect' ratio).

One of the up sides of 7z is that *if you have lots of memory* you can
usually squeeze a few extra bytes by using insane parameters. However,
this gets very slow (especially if it starts swapping!), and may
require quite some memory for decompression too. (The manual explains
the exact requirements.)

-- 
Giuseppe "Oblomov" Bilotta

"Da grande lotterò per la pace"
"A me me la compra il mio babbo"
(Altan)
("When I grow up, I will fight for peace"
 "I'll have my daddy buy it for me")