A side-effect of this is that any description of a stars block that has one-byte-or-smaller pieces may be incorrect. For instance, alignment of the FileHeaderBlock here:

http://wiki.starsautohost.org/wiki/FileHeaderBlock

probably has bytes 16 and 17 switched. This also clarifies instances where there is weirdly packed data, such as in places like the block header shown here:

http://wiki.starsautohost.org/wiki/Technical_Information#Gen eral_File_Structure

I'm thinking that the file byte stream should be read in swapped to begin with - but this may change paradigms a little. It will definitely reduce deserialization code.