Igor, do you plan to support/improve such new compression algorithm and WIM format in 7-zip?
As you can read in wimlib 1.6.0 documentation (--> https://sourceforge.net/projects/wimlib/ ), latest DISM tool from Microsoft can export images from old WIMs with /compress:recovery and pack them with this new format, which supports: 1) solid blocks; 2) new LZMS compression algorithm (similar to LZMA).
OK. I'll check it later. Thanks for that information!
Igor, I don't think this feature is high priority for 7-Zip since MS's own software (ImageX, Dism) has problems with the new WIM format, and there is no official documentation. However, just a few notes if you're interested:
Dictionary size for LZMS can be up to 67108864 bytes.
LZMS uses range coding from the beginning of the block and Huffman coding from the end of the block. For matches it uses a queue of 3 recent offsets, and it also has a weird match type which I don't fully understand yet. It's an interesting format but seemingly isn't as good as LZMA, overall.
I haven't had much luck (compatibility-wise) using algorithms other than LZMS in solid blocks, or using multiple solid blocks per WIM archive.
LZMS is undocumented (even though it is one of the algorithms supported by the public Compression API available on Windows 8), so if you ever want to support it in 7-Zip it might be best if I wrote up a formal spec and you went from there rather than looking at my code, especially as I did some disassembling to get started in the right direction. Maybe MS will release a spec eventually like they did with XPRESS and LZX though.
Log in to post a comment.