in binary mode and comparing the content to strings,
which failed. Now the bytes get decoded into unicode
using latin-1 (the comparison compares ASCII strings
only anyway, and we don't want the decoding to fail).