This appears to be an issue with running against python 3.x

in Python 2.x, base64.b64encode takes a string object

in Python 3.x, it takes a bytes object

I'm not sure what the best approach is for making it compatible with both, but here's how I got it working in python 3.x:

    return {
        'iam_http_request_method': request.method,
        'iam_request_url':         base64.b64encode(bytes(request.url, 'utf-8')),
        'iam_request_body':        base64.b64encode(bytes(request.body, 'utf-8')),
        'iam_request_headers':     base64.b64encode(json.dumps(headers_to_go_style(dict(request.headers)))), # It's a CaseInsensitiveDict, which is not JSON-serializable
    }

also had to change iteritems to items

def headers_to_go_style(headers):
    retval = {}
    for k, v in headers.items():
        retval[k] = [v]
    return retval

probably some more changes too... I'll log them here as I find them

I'm not sure what the best approach is for making it compatible with both

I think it's okay if it's py 3 only. PRs welcome! :)

What about adding a separate py3 version of the file (sign-request-py3.py)? Then users can pick and choose which one to include in their AMI when they are copying the examples.

I agree that we shouldn't try to make the existing file py2 + 3 compatible because without six, that will make the code littered with if statements.

okay, it needed a few more tweaks, but here's a version for 3.x:

https://gist.github.com/eedwards-sk/161e98844db85919f6a85a3f7dce58f7

please feel free to toss into a PR or we could use this to figure out a 3.x/2.x compatible version