If this issue has raised previously, then I am sorry for repeating.  I did a search but did not find related reports.

Below is the thing I did.

  config = configparser.RawConfigParser()
  #config.read_file(urlopen(path_config))
  config.read_file(open('jkl.ini', 'rb'))

The line commented out was the thing I wanted to do originally.  I wanted to parse a configuration file stored on some web server.  And I got this error "TypeError: startswith first arg must be bytes or a tuple of bytes, not str."  But after I tried, with this line "config.read_file(open('jkl.ini', 'rb'))", the same error can be reproduced.  Therefore, I think the error message should be stated another way around as "startswith first arg must be str instead of bytes or a tuple of bytes."  I have checked this by adding the lines below to configparser.py after the for-loop at line 994.

  print(type(line))
  line = str(line, 'utf-8')
  print(type(line))

That made the code work.

Hello, Mickey. By doing open('file', 'rb') you're explicitly stating you want the file to be opened in BINARY mode which means it doesn't return strings but bytes. This is not supported anymore in Python 3. This is clearly documented here: http://docs.python.org/dev/library/configparser.html#configparser.ConfigParser.read_file

Same goes for urllib.request.urlopen, it returns bytes because this is all Python knows at the time of reading the URL. If you happen to know the encoding (for instance by parsing the data or headers) you can do something like this: http://docs.python.org/dev/library/urllib.request.html#examples . Your example would become:

config.read_string(urlopen(path_config).read().decode('utf-8'))

Since this is a reasonable mistake to make, I'm leaving this open to adjust the exception raised if read() yields bytes instead of expected strings.

Mickey, you can wrap file-like object returned by urlopen with io.TextIOWrapper.

  config = configparser.RawConfigParser()
  config.read_file(io.TextIOWrapper(urlopen(path_config), encoding='utf-8'))

Because there is no bug and new feature is not needed, I believe that this issue can be closed.