KGB Archiver

[pawnara]

Hi there! I have a suggestion regarding one packer, namely it is a KGB Archiver ( www.kgbarchiver.net ) compressor which compresses files far better than any other out there.
The program is free and the source code is public so implementing it shouldn't be a problem (except for the difference in languages but I haven't yet seen in which it is programed). I hope you will consider this seriously.

Regards.

[fenix_productions]

2pawnara
I don't really think that it will be implemented internally. Even much more popular 7zip is not.

I don't even think that someone will write a plugin for this. Why? Monstrous memory usage and compression duration are the answers.

[pawnara]

Here are two simple test:

Code: Select all

Test 1
Here I am using a 10kb file filled with spaces.

rar  (max)  -  98b   / below 10s
tgz  (max)  -  119b / below 10s
zip  (max)  -  150b / below 10s
gz   (max)  -  59b  / below 10s
kgb (4/10)  - 38b   / 11s

Code: Select all

Test 2
I am using VB6 exe (original size 1 880 064b):

rar (max) -    866 419b / below 10s
tgz (max) - 1 019 515b / below 10s
zip (max) - 1 019 594b / below 10s
gz  (max) - 1 019 508b / below 10s
kgb (4/10) -   806 058b / 30s
kgb (8/10) -   698 260b / 4min

The conclusion is: kgb takes much more time to compress but the results are far more than impressive. These days more and more people are upgrading their machines hence kgb memory requirements are not a big deal, some people may like to have their files ultra small and when they can do it with their favourite file manager, that would be heaven.

[ghisler(Author)]

Could you add 7zip and bzip2 to the comparison, please?

[Hacker]

KGB is based on the sources of PAQ6 which is now more than 4 years old. The PAQ algorithm is in constant experimental development causing backwards incompatibility, so it does not make sense to simply pick out a 4-year old version (or, rather, its specific implementation in the KGB archiver) and support that.

And there are certainly other packers that compress better than KGB, like the abovementioned newer PAQ implementations, or WinRK using PWCM (which is fully closed, so no chance to support).

On the other hand, I would certainly welcome internal 7-Zip support using the 7za.dll which is LGPL'd.

References:
PAQ on Wikipedia
Maximum Compression comparison test
7-Zip FAQ

Roman

[fenix_productions]

2pawnara
You are writing about "upgrading machines" and trying to show small files compression. Who does that?
If I have thousands of small files then more important for me is to have them in one archive (smaller fragmentation) so size doesn't really matter. And for the big files: more important is the time. TC doesn't support packing / unpacking in background for plugins so it is important to answer: how long should I wait or how many TC instances should I use when I want to do simple backup. In this case: zip wins (or maybe external WinRAR).

Other important thing is popularity. I had started using computer back in 1995 (learning on friends PC). Later I've got my own and up to now I've seen KGB archives few times and only because I've wanted to test it. IMHO much more popular ACE or LHA are. It is undeniable that ZIP is the winner for popularity. The second place might be occupied by RAR. Other formats are too rare for internal implementation. Only 7zip might stand a chance because of Internet growth and the people who host their files this way.

If there is a need for new format internal support than I will vote for 7zip.

[ghisler(Author)]

Yes, I had similar thoughts, that's why I asked for comparison with 7zip. Sorry, don't have VB, so I can't test it myself.

[m^2]

Compression slowness is not the worst thing about PAQ. Unlike with most algorithms, decompression isn't any faster.

[MVV]

7-Zip grants great compression. And it grants a much faster decompression than RAR (of cource, I mean solid archives).
It will be great if internal support of 7-Zip will be implemented.