CHANGELOG

Version ?:
* Add support for connect_options
    * Enables the user to pass socket options, for instance ip and port,
      that will be used when connecting the socket
    * Allows the user to specify SSL options during the connect phase
* Add support for "streaming" of entities
* Add start/0 and stop/0
* Fix for unexpected messages after request has been completed
    * When the client process is trapping exits (which some eunit versions seem
      to be doing by default) there would be {'EXIT', Pid, normal} messages left
      after completing requests which came from the lhttpc_client process.
      These are now avoided.
* Add rebar support (thanks to Benoit Chesneau)
* Add support for HTTP basic auth (Filipe Manana)
* Add support for IPv6 (Filipe Manana)
* Several URL parsing fixes (Filipe Manana)
* Add support for proxies via HTTP CONNECT method (Filipe Manana)
* Add connection limit to the connection pool (Filipe Manana)
* Add suport for using multiple connection pools - different requests can now
  use different connection pools (Filipe Manana)

Version 1.2.5:
* Fix for decoding chunked HTTP responses with extraneous whitespace
  (thanks to Bob Ippolito)
    * api.facebook.com includes extra whitespace in its chunked HTTP response
      length, e.g. "6  \r\n" instead of "6\r\n". 

Version 1.2.4:
* Fix handling of HTTP response code 304 (patch by Ben Slavin)

Version 1.2.3:
* Fix handling of unexpected 100 Continue response (patch by Magnus Henoch)
 
Version 1.2.2:
* Fix Host header, add port number (reported by Benoit Chesneau)

Version 1.2.1:
* Fix confusion of arguments in request/9 (introduced in 1.2.0)

Version 1.2.0:
* Add support for options
    * {connect_timeout, Milliseconds}: Aborts the connect phase after
      Milliseconds
    * {send_retry, N}: The client will retry sending the request N times
      if the connection is closed
* Add support for downloading chunked transfer encoding
* More consistent error handling
    * The client can either return {ok, Result} or {error, Reason}. Reason
      (which wasn't very well defined before) can now be:
        * connection_closed: The server closed the connection on us (N times
          in case it happens just after sending the request) 
        * connect_timeout: If the TCP stack gives up or we hit the
          connect_timeout option
        * timeout: If the overall request timeout value is hit
    * All other errors (socket, protocol etc. will result in a runtime error)
* Better connection handling (issues #2 and #3 on bitbucket.org/etc/lhttpc)
    * Now understands what to do with connections to servers < HTTP/1.1
    * Now respects "Connection: close" in request headers, which means
      clients can choose to not use persistent connections
* RFC Compliance
    * Fix reading of chunked encoding since section 3.6 claims that "All
      transfer-coding values are case-insensitive"
    * Support for responses that doesn't have an entity body (thanks to Steve
      Ellis)
        * No body for the HEAD and OPTIONS (unless indicated for the latter)
          methods
        * Don't try to read a body when receiving a 204 status

Version 1.1.2:
* Fix minor error with {active, once} instead of {active, true} in manager
* Remove socket and try to find another if the manager can't set controlling
  process due to socket error
* Improve test suite 

Version 1.1.1:
* Fix problem with empty lists in dicts when last socket is consumed
* Improve test suite to include cover report 

Version 1.1.0:
* Support for configurable connection timeout 

Version 1.0.0:
* Initial version
* Persistent connections have hardcoded timeout