[rabbitmq-discuss] Subscription Pattern in .net

[Emile Joubert]

Hi Steven,

On 15/03/11 20:31, Steven Taylor wrote:
> This
> is slightly unrelated, but it looks to me as though there are a few
> rough edges we could sand back.  e.g. channel has a
> "CreateBasicProperties" method that we can use in BasicPublish.  No big
> deal, but BasicPublish could have an overload which creates its own
> properties, as this method knows its context already (i.e. the method is
> in Channel, so go create your own basic properties, basicPublish!!)

That would be similar to supplying null properties. So far there has 
been insufficient demand for an overloaded publish method, because it is 
simple enough to supply null properties.

> No reset required for the enumerator.

I agree that it makes no sense to reset a subscription, but the 
enumerator is useless after MoveNext returns false without a Reset(), 
according to the IEnumerator documentation.

> The context is: A. send off a bunch of messages asynchronously, and B.
> wait for a reply.  C. If messages don't turn up in a reasonable period
> of time, stop waiting.

What is the state of the enumerator at this point? Should it be 
considered "irrecoverably invalidated" as the docs suggests? Does it 
make sense for a gap between messages to invalidate the enumeration in 
this way?

Your use-case is not general enough justify changes to the Subscription 
class. If there is sufficient demand then we will gladly reconsider your 
proposal, but perhaps under a different name or in a different class.

> However, being in a
> community that has helped out, I thought, hmmm... maybe i could at least
> return the favour in a small way.

Thanks for the feedback. We appreciate the suggestions. It is through 
suggestions like yours that rabbit continues to evolve.



Regards

Emile