As a friend above mentioned, it is impossible to compress all files in a lossless manner. It doesn't even need Shannons theorems to prove this. Nonetheless, many files can be compressed thanks to the redundancy in their structure. The limmit to the compressibility of a file is given by it's Kolmogorov Complexity K. Unfortunatly, K is not a computable number (this is also proved by Kolmogrove himself) so this ultimate compression can not be achieved in practice.

Li and Vitanyi have a very beautiful book on this topic:

http://homepages.cwi.nl/~paulv/kolmogorov.html

cheers

Although this is an old topic,..
I think it's just a manipulation of the field "Uncompressed size".
Is it rigth dear Ehsan?

The rar files structure

Type Value

OFFSET LABEL TYP VALUE DESCRIPTION
------ ----------- ---- ----------- ----------------------------------

00 HEAD_CRC DW 0000 Low part of the CRC header
02 HEAD_TYPE DB 00 Block type

Value: 72h = Marker block 73h = Archive header
74h = File header 75h = Comment header
76h = extra information 77h = subblock
78h = recovery record

03 HEAD_FLAGS DW 0000 Block size

Value: 4000h = old archive type , Ignored by older version
8000h = if set, ADD_SIZE field is present

05 (ADDSIZE) DW 0000 Optional - added block if file block

07 RARCOMPSS HEX 00000000 Compressed size
0B RARUNCMPS HEX 00000000 Uncompressed size
0F HOSTOS DB 00 Host operating system

Value: 0 = MS DOS 2 - WIN32
1 = OS/2 3 - UNIX

10 CRC32 HEX 00000000 CRC32 of the original file
14 FILETIME DW 0000 File time , msdos format
16 FILEDATE DW 0000 File date , msdos format
18 VERNEED DB 0 Version need to extract
19 METHODCMP DB 0 Compression method

Value: 0x30 - storing 0x33 - normal compression
0x31 - fastest compression 0x34 - good compression
0x32 - fast compression 0x35 - best compression

1A NAMELENGTH DB 0 File name length
1B ATTRIB DW 0 Rar attribute
1D FILENAME DS ? Filename[ltr]