在读如下这篇文章之前，对消息队列尚有几个问题不甚明朗。当然，读完该文后，关于队列的概念就一目了然。
http://rubyamqp.info/articles/working_with_queues/
我在RabbitMQ邮件列表上问了几个问题：
- 获取消息时，consume和get方法的区别是什么？
- 如何设置消息ack？
- 假如消息ack打开了，comsumer是自动发送ACK，还是要手工调用方法发送？
- 如何设置消息事务，例如提交和回滚消息？
来自RabbitMQ开发公司的Emile Joubert回答了这几个问题，贴出来共享之。
> * What’s the difference between “consume” and “get” for a consumer?
The one is asynchronous and the other synchronous. When a client uses
“consume” the broker will deliver messages as soon as any are available,
while with “get” the client needs to poll the broker.
> * Where to setup the message acknowledgment?
Clients decide on this via the “no-ack” parameter of the “basic.consume”
or “basic.get” methods.
http://www.rabbitmq.com/amqp-0-9-1-reference.html#basic.consume.no-ack
http://www.rabbitmq.com/amqp-0-9-1-reference.html#basic.get.no-ack
> * If message acknowledgment is enabled, will the consumer send the ACK
> automatically, or by a method call?
If the consumer set “no-ack/auto-ack” true then the broker will not
expect acknowledgements from the consumer.
If the consumer set “no-ack/auto-ack” false then the consumer must
acknowledge messages using the basic.ack method:
http://www.rabbitmq.com/amqp-0-9-1-quickref.html#basic.ack
Also see section “2.1.2.1 Message Life-cycle” in the 0.9.1 AMQP
specification for a more detailed description.
> * Any guide for transaction? for example, commit and rollback a transaction.
Transactions are discouraged, but see this for more details:
http://www.rabbitmq.com/semantics.html#tx
http://www.rabbitmq.com/amqp-0-9-1-quickref.html#class.tx
Publisher confirms are recommended instead of transactions:
http://www.rabbitmq.com/extensions.html#publishing