[] notation for IPv6 with X-Forwarded-For

[NicolasLM]

After upgrading to Waitress 1.2.0 and letting it handle X-Forwarded-For headers, the WSGI environment REMOTE_ADDR is populated with IPv6 using the [] notation, for instance '[2a02::1]'.

While the WSGI specification doesn't say anything about REMOTE_ADDR, it is a bit surprising since this notation cannot be used with ipaddress:

>>> import ipaddress
>>> ipaddress.ip_address('2a02::1')
IPv6Address('2a02::1')
>>> ipaddress.ip_address('[2a02::1]')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.4/ipaddress.py", line 54, in ip_address
    address)
ValueError: '[2a02::1]' does not appear to be an IPv4 or IPv6 address

[digitalresistor]

TL;DR: The [] needs to be stripped from the IPv6 address, with the caveat that at that point you can't easily do env['REMOTE_ADDR'] + ':' + env['REMOTE_PORT'] and get a "nice" representation

Background:

The REMOTE_ADDR is actually from the CGI spec which the WSGI specification kind of inherits from.

The CGI spec: https://tools.ietf.org/html/rfc3875#section-4.1.8 mentions https://tools.ietf.org/html/rfc3513#section-2.2 which is the standard for IPv6. However https://tools.ietf.org/html/rfc4291#section-2.2 obsoletes RFC3513. Then there is errata for RFC4291 that is https://tools.ietf.org/html/rfc5952#page-11 which shows how to best do text presentation of IPv6.

When we get the X-Forwarded-For header, it may include the remote port, in this format:

X-Forwarded-For: [2001:db8::1]:30943

We peel off the 30943 and set that as the REMOTE_PORT and the rest is REMOTE_ADDR. This way if you take env['REMOTE_ADDR'] + ':' + env['REMOTE_PORT'] you get back [2001:db8::1]:30943, if we strip [] from the IPv6 address, you'd end up with 2001:db8::1:30943 which while a valid way to represent an IPv6 address + port (RFC5952) is not very readable and requires extra parsing to put together the real remote pair.

I would argue that ipaddress should be updated to support bracketed IPv6 as well, but it also doesn't support scoped IPv6 addresses either, so there's some other needs too.

[digitalresistor]

@NicolasLM I fixed and released version 1.2.1

[NicolasLM]

Thank you for your work on this!