I'm curious about what happens to string with non-latin1 characters in them:
do they cause the loading to abort, or are they interned as string
with different lengths
depending on the unicode support? (Similarly for latin1 strings that
are malformed as utf-8.)