Use urllib3 connection pooling to improve remote server connection
speed. Our aim is to reuse socket connections to the same target hosts when
possible.
Initialize a `urllib3.PoolManager` in `SingleThreadedWarcProxy` and use
it in `MitmProxyHandler` to connect to remote servers.
Socket read / write and ssl / socks code is exactly the same, only the
connection management changes.
Use arbitratry settings: pool_size=2000 and maxsize=100 (number of
connections per host) for now. Maybe we can come up with better values in the
future.
`certauth` has a method to create a cert for `*.example.com`. This
reduces greatly the number of generated certificates (~50% in my
tests).
For example, previous code would create:
```
images-eu.ssl-images-amazon.com.pem
images-fe.ssl-images-amazon.com.pem
images-na.ssl-images-amazon.com.pem
```
Wildcard code would create:
```
ssl-images-amazon.com.pem
```
We use the default list of SSL ciphers of python `ssl` module when we connect
to remote hosts. That list is probably outdated.
https://github.com/python/cpython/blob/3.6/Lib/ssl.py#L192
We noticed problems when connection to various targets. E.g.
```
2018-01-31 21:29:23,870 3067 WARNING
MitmProxyHandler(tid=8052,started=2018-01-31T21:29:22.501118,client=127.0.0.1:56340)
warcprox.warcprox.WarcProxyHandler.log_error(mitmproxy.py:447) code 500,
message EOF occurred in violation of protocol (_ssl.c:645)
2018-01-31 21:29:23,987 3067 ERROR
MitmProxyHandler(tid=7327,started=2018-01-31T21:29:22.741262,client=127.0.0.1:56448)
warcprox.warcprox.WarcProxyHandler.do_CONNECT(mitmproxy.py:311) problem
handling 'CONNECT beacon.krxd.net:443 HTTP/1.1': SSLEOFError(8, 'EOF
occurred in violation of protocol (_ssl.c:645)')
2018-01-31 21:29:23,870 3067 ERROR
MitmProxyHandler(tid=8052,started=2018-01-31T21:29:22.501118,client=127.0.0.1:56340)
warcprox.warcprox.WarcProxyH
andler.do_CONNECT(mitmproxy.py:311) problem handling 'CONNECT
px.surveywall-api.survata.com:443 HTTP/1.1': SSLEOFError(8, 'EOF
occurred in violation
of protocol (_ssl.c:645)')
```
Research indicated that the cipher selection is not proper.
I use `urllib3` cipher selection for better compatibility.
https://github.com/shazow/urllib3/blob/master/urllib3/util/ssl_.py#L71
The `urllib3` list is bigger and includes TLS13 which from my experience
is the latest state of the art.
`ssl` module ciphers:
```
'ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:ECDH+HIGH:DH+HIGH:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+HIGH:RSA+3DES:ECDH+RC4:DH+RC4:RSA+RC4:!aNULL:!eNULL:!MD5'
```
`urllib3` module ciphers:
```
'TLS13-AES-256-GCM-SHA384:TLS13-CHACHA20-POLY1305-SHA256:TLS13-AES-128-GCM-SHA256:ECDH+AESGCM:ECDH+CHACHA20:DH+AESGCM:DH+CHACHA20:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:RSA+AESGCM:RSA+AES:!aNULL:!eNULL:!MD5'
```
Add ``ProxyingRecordingHTTPResponse.truncated`` and
``RecordedUrl.truncated`` attributes.
Set ``truncated = b'length'` when max resource size limit applies.
Add `Warc-Truncated: length` header to WARC record.
Add hidden option ``--max-resource-size`` which indicates the max file size of
a target resource in bytes. If the size is over the limit, an exception is
raised.
When the option is not set, use existing single threader writer
architecture.
If available, load ``WarcWriterMultiThread`` with pool size equal to
``--writer-threads``.
Add socket-timeout=4 in ``warcprox_`` test fixture.
Create test URL `/slow-url` which returns after 6 sec.
Trying to access the target URL raises a ``socket.timeout`` and returns
HTTP status 502.
The new ``--socket-timeout`` option does not hurt any other test using
the ``warcprox_`` fixture because they are too fast anyway.