Release: RabbitMQ 1.2.0
Status : beta

Release Highlights
==================

RabbitMQ server
---------------
- introduced internal flow control to prevent performance drops when
  running a server near capacity
- simplified cluster configuration and added "automatic" clustering
- made rabbitmqctl command line syntax less confusing
- fixed a couple of race conditions that could result in a client
  receiving unexpected sequences of command replies and messages
- refactored internals to make it easier to construct handlers for
  transports other than raw TCP/IP

Java client
-----------
- fixed a race condition between invocation of a Consumer's
  handle{Consume,Cancel}Ok and handleDelivery methods; the former are
  now called inside the connection's thread, just like the latter

Packaging
---------
- simplified rabbitmqctl invocation under Debian
- moved default location of the log and mnesia dirs under Windows to
  sub directories of the RABBITMQ_BASE directory
- changed startup scripts to allow the rabbitmq_server package to
  reside outside the OTP library directory


Upgrading
=========

Upgrading to this release from RabbitMQ 1.1.x requires no special
steps at the server end. There have been some minor changes to the
Java client API. Most client code is unlikely to be affected by these,
and is easily changed if it is.

If you are upgrading from an earlier release of RabbitMQ, note that
RabbitMQ's internal database schema has changed, and hence the
MNESIA_BASE directory should be cleared before starting the new
server. This wipes out any configuration information and persisted
messages.


Credits
=======

We would like to thank the following individuals for submitting bug
reports and feedback that we incorporated into this release:

Ben Hood
Emmanuel Okyere
Holger Hoffstätte
Jodi Moran
Robert Greig