md5sum Hard drive vs CD-ROM puzzle

[Ron Stordahl]

In order to burn the 2 CD i386 binary set of 2.1r3 I ftp'd the images to my Win98 machine as I don't currently have a CD burner in my Linux box.  I have successfully used these CD's to create Linux systems.  These CD's were burned using EZ CD Creator using image mode.  But because I have limited storage on my Win98 machine I also burned the image files, binary-i386-1.iso and binary-i386-2.iso in file mode to two CD's so that I could delete them from my hard drive and restore them later should I wish.  After burning them to the CD's I did a byte by byte comparison of the CD's vs the disk files on my Win98 machine using DOS command fc/b and found them to be identical.   Being properly paranoid I wanted a 'second opinion' which I could get by running md5sum on the files.  The md5sum is provided on the various mirror sites for the disk images.  I needed to get the files to my Linux machine to run md5sum.    So I took the first CD, which contained the file binary-i386-1.iso and copied is from my CD reader on my Linux machine to my home directory on my hard drive.  I ran md5sum on the hard drive file and obtain the correct checksum, as is reported in the file MD5SUMS.txt.  I am thus confident that the file I now have on my hard drive is identical to the file I ftp'd from the mirror site to my Win98 machine, burned on a CD-ROM, carried to my Linux machine etc...    But when I run md5sum against the same file on the CD-ROM I get a different md5sum!!   So from my home directory I ran cmp binary-i386-1.iso /cdrom/binary-i386-1.iso and no differences are found between the files.   I have repeated this process several times, its entirely reproducible.   Two identical files produce different md5sums!  How can this be?   Ron