Added tag 1.0.2 for changeset 124bd71678e2

Bump the version number to 1.0.2 for the packing fix for Windows

Fix Issue 42 having to do with installing on Windows

Oops, forgot to document what was changed in 1.0.1

Added tag 1.0.1 for changeset cf8638fcc5c1

Fix Issue 41 relating to a minor problem some older versions of FreeBSD have with socket.getaddrinfo()

Add MANIFEST.in, so that demos, tests, docs are included in the
source distribution. Fixes Issue 36 , thanks to MichaelF...@gmail.com
for the report and the patch

Add more Trove classifiers to help packagers, bump the version number for the next release

Added tag 1.0.0 for changeset 59dc52a3c871

Wrap this thing up, call it 1.0.0 in celebration of OSCON 2011

Fix Issue 35 , comparing a Message to something else (such as None), raises an AttributeError
instead of returning false. Thanks to Tarrant....@gmail.com for reporting this.

Process basic_return methods from the server whenever they're received, rather than queueing them up and basically ignoring them unless the user called wait(). (Issue 29, thanks to pkwar...@gmail.com for bringing up the issue and providing a patch)

Enable TCP keepalive (Issue 32, suggested by g...@theoldmonk.net)

Enable TCP_NODELAY, gives a nice little 6x speedup of the unittests at least.

Shutdown a socket before closing it, to ensure pending transmissions aren't discarded. (Issue 30, thanks to chr...@pianocktail.org for pointing this out).

Add support for IPv6, supporting literal addresses or domain names with AAAA records.

Add support for queue_unbind, which is supported by RabbitMQ as an extension to the 0-8 protocol.
Thanks to Hasan Alayli <halayli@...> for a patch.

Alter one of the unittests skipped over for Python 3.x so that it can run and check the behavior we want even in 3.x

Add patch from Viktor Naoumov <vnaumov@r...> adding the ability on Python >= 2.6 to specify a dictionary of SSL options to pass to ssl.wrap_socket(), sock as specifying a client cert, a CA cert, and requiring the ca cert match.

Since encoding of unicode message bodies has now moved from happening at Message creation time to transmission time, move the setting of a default content_encoding message property to happen at the same time. So they're both late behaviors now, instead of one being early and one being late.