feature request: lzma compresion (7zip)
Matthias Schniedermeyer
ms at citd.de
Mon Jun 21 04:16:26 MDT 2010
On 21.06.2010 10:53, Matthias Schniedermeyer wrote:
> On 17.06.2010 15:37, super master wrote:
> > Hello,
> > many linux SW is starting to implement new lzma compresson instrad of old zlib (gzip) od bzip2.
> >
> > lzma is default comrpession in very good compression SW 7-zip, which is faster and have higher compression ratio then bzip2 or rar.
> >
> > Currently its probalby the best compressor in therms of compression and decompression speed / compression ratio.
> >
> > In linux there is GNU lzma SW which implements lzma algorithm.
> >
> > In my optinion it would be very useful if rsync could use this compressor optionaly as upgrade of old zlib.
> >
> > more info: http://en.wikipedia.org/wiki/Lempel%E2%80%93Ziv%E2%80%93Markov_chain_algorithm
>
> lzma draws most of it's strength from using a much larger "Working
> Window" (megabyte to gigabyte, instead of a few kilobyte like gzip).
>
> AFAIU rsync works with blocks in the kilobyte-range, which effectivly
> prevents lzma from using it's main strength.
>
>
> For illustration i've just tried "xz -0" on a linux-kernel tar-ball.
> "xz -0" still uses a much larger window that gzip! The man-page says
> about 6 megabytes and the man-page reserves a future change to an
> algorithm other than lzma.
>
> On said linux-kernel tar-ball the compression was only slightly better
> that gzip.
>
> 87634801 byte for gzip with defaults
> 86841598 byte for gzip -9
> 85544824 byte for xz -0
>
> For comparison:
> 55044372 byte for xz -e9
A little addendum.
I have looked on the wrong table in the man-page. memory-usage vs.
dictionary size.
I used the first, while the other is the correct one.
For "xz -0" the dictionary size of 64KiB is "only" twice the maximum
dictionary size of deflate and not orders of magnitute larger than the
"64MiB" for -9.
Bis denn
--
Real Programmers consider "what you see is what you get" to be just as
bad a concept in Text Editors as it is in women. No, the Real Programmer
wants a "you asked for it, you got it" text editor -- complicated,
cryptic, powerful, unforgiving, dangerous.
More information about the rsync
mailing list