What about first line? Currently both Python interpreter and the tokenize module decode it from UTF-8 (actually due to bug #18960 Python interprets it twice, in different encodings). PEP 263 says:

    1. The complete Python source file should use a single encoding.
       Embedding of differently encoded data is not allowed and will
       result in a decoding error during compilation of the Python
       source code.

I conclude that the first line should be decoded with the encoding specified in the second line. We first should read the first line, check if it isn't a comment or contains encoding cookie, otherwise read the second line, determine the encoding, and decode read lines. Perhaps it will untangle issue18960 too.