Author Topic: A suggestion on distribution  (Read 9357 times)

ykuksenko

  • Guest
A suggestion on distribution
« on: July 22, 2005, 03:51:30 pm »
I would like to suggest to those that host the download files, please try the 7-zip format to compress those downloads they will compress much better!!!

Offline rickg22

  • Lives here!
  • ****
  • Posts: 2283
A suggestion on distribution
« Reply #1 on: July 22, 2005, 05:59:39 pm »
Thanks for the suggestion, but there's a problem... CodeBlocks is intended for a broad audience. They usually have winzip in their machines, and making them install 7zip would be an unnecessary annoyance. We may lose quite a few users if we do that.

Offline kagerato

  • Multiple posting newcomer
  • *
  • Posts: 56
    • kagerato.net
Re: A suggestion on distribution
« Reply #2 on: July 22, 2005, 10:47:30 pm »
Quote from: ykuksenko
I would like to suggest to those that host the download files, please try the 7-zip format to compress those downloads they will compress much better!!!


I suggest quantifying your claim with some actual data.  Otherwise, it is difficult to determine whether "much better" is a significant difference.

darklordsatan

  • Guest
A suggestion on distribution
« Reply #3 on: July 22, 2005, 11:06:54 pm »
Quote from: kagerato

I suggest quantifying your claim with some actual data. Otherwise, it is difficult to determine whether "much better" is a significant difference.


I think is not a big secret 7-zip format is the big daddy of compression formats right now...


Compression ratio

Compression ratio results are very dependent upon the data used for the tests. We compared 7-Zip with some of the leading archivers: WinRAR 3.10, WinAce 2.3, CABARC 1.0, PKZIP 2.50.

FILE SET: The GIMP 1.2.4 for Windows after full installation (127 subfolders, 1304 files totaling 27,128,826 bytes). The GIMP is the GNU Image Manipulation Program. It can be downloaded from www.gimp.org.
Code

Archiver     Compressed size Ratio
7-Zip (7z format)    5445402           100%
WinRAR 3.10           6004155          110%
WinAce 2.3             6242424           115%
CABARC 1.0           6455327           119%
7-Zip (zip format)    9461621          174%
PKZIP 2.50             9842800           181%


Usually 7-Zip compresses to 7z format 30-70% better than to zip format. And 7-Zip compresses to zip format 2-10% better than other zip compatible programs.

Offline rickg22

  • Lives here!
  • ****
  • Posts: 2283
A suggestion on distribution
« Reply #4 on: July 23, 2005, 12:20:26 am »
Did you include BZIP2 in comparison? (slightly offtopic, bzip2 would have rocked except for its ridiculous one-file-only approach. Why didn't they just use the infozip format for storing the headers?)

darklordsatan

  • Guest
A suggestion on distribution
« Reply #5 on: July 23, 2005, 12:44:12 am »
IMHO, 7-zip usually owns bzip2, but the later is also excellent. I just havent make a comparison between the two of them yet, but for general information, from all the "built-in" linux (de)compressing tools, its the best one.

Offline rickg22

  • Lives here!
  • ****
  • Posts: 2283
A suggestion on distribution
« Reply #6 on: July 23, 2005, 12:57:22 am »
I studied data compression, and bzip uses suffix sorting (very interesting. Search google for PPM compression algorithms and you'll see). I don't know the algorithm for 7-zip, but i suppose it's similar.

In any case, here's some Google info on popularity for the three formats:

about 115,000,000 for zip -"7-zip" -bzip -bzip2
about 2,700,000 for "7-zip" + 377,000 for 7zip
about 933,000 for bzip2 + 236,000 for bzip

See, obviously, ZIP has the advantage in popularity. :. Q.E.D.

Offline kagerato

  • Multiple posting newcomer
  • *
  • Posts: 56
    • kagerato.net
A suggestion on distribution
« Reply #7 on: July 23, 2005, 03:58:02 am »
Quote from: darklordsatan
I think is not a big secret 7-zip format is the big daddy of compression formats right now...


Certainly there are quite a few cases where it offers superior compression (though not necessarily better speed, or speed-to-compression ratio) than other available reduction algorithms.  That statement is especially relevant concerning PKZip because of its age.

However, it is unwise to consider alternatives without actual comparison data.  When one is missing actual figures, it is easy to be lured into the impression of great gains or improvements by generalized statements.  The actual magnitude of difference is rarely as great as proponents would lead you to believe.

Quote from: rickg22
See, obviously, ZIP has the advantage in popularity. :. Q.E.D.


This is accurate, but your google results are skewed in magnitude.  The word zip receives a good deal more results because it has multiple meanings and more potential contexts.  I can practically guarantee that if you sorted through those 115 million results, many of them would show nothing relating to compression.

Offline thomas

  • Administrator
  • Lives here!
  • *****
  • Posts: 3979
Re: A suggestion on distribution
« Reply #8 on: August 05, 2005, 10:49:46 pm »
Actually, 7zip does not only compress significantly better and is a LOT faster than bzip (as far as decompression is concerned, compression is really painful, 2-3 times slower sometimes). LZMA, the algorithm used in 7zip is an optimized sliding window compressor, so unlike bzip, decompression does not need to do an awful lot except a few memmoves (no inverse transform calculation etc).

Since you compress once users only decompress, this sounds perfect. However, as stated before, the problem is not one of good technical stats, it is one of availability.
A lot of people will refuse to install a program to decompress the package of another program (and, somehow rightfully, too). To make matters worse, the Windows client coming with 7zip is quite ugly. Not that is doesn't work, it is just ugly.

However, speaking of LZMA and availability... what installer does code::blocks actually use?
NSIS, for example, has transparent support for LZMA built in.
"We should forget about small efficiencies, say about 97% of the time: Premature quotation is the root of public humiliation."

darklordsatan

  • Guest
Re: A suggestion on distribution
« Reply #9 on: August 05, 2005, 11:08:42 pm »
what installer does code::blocks actually use?
It uses Inno Setup

Offline thomas

  • Administrator
  • Lives here!
  • *****
  • Posts: 3979
Re: A suggestion on distribution
« Reply #10 on: August 06, 2005, 01:50:35 pm »
And indeed, I looked at CodeBlocks_Core.iss and CodeBlocks_MINGW.iss, the installer already uses LZMA. :)
Guess that concludes our discussion.
"We should forget about small efficiencies, say about 97% of the time: Premature quotation is the root of public humiliation."