The only other thing I can think of is to have two types of consumers, 'A' and 'B,<br>with corresponding exchanges, and send messages out once to each of the<br>exchanges. This is a suboptimal because you lose any global load balancing<br>
or fault tolerance. It could also be done with header values.<br><br>That's all I've got.<br><br>I am curious as to why 2 consumers. Is it that the processing needs to be<br>verified, is there a trust issue, or something else? If the reason is generally<br>
applicable, then maybe a feature should be added to AMQP to support it.<br><br>- Jim<br>
<br><br><div class="gmail_quote">On Fri, Sep 4, 2009 at 10:47 AM, Benoit Chesneau <span dir="ltr"><<a href="mailto:bchesneau@gmail.com">bchesneau@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">On Fri, Sep 4, 2009 at 4:38 PM, Jim Irrer<<a href="mailto:irrer@umich.edu">irrer@umich.edu</a>> wrote:<br>
> You could use a direct queue and send the message twice, which<br>
> is simple but incurs a little extra overhead.<br>
</div>The problem here is that I could have a lot of nodes behind. But this<br>
solution seem simple.<br>
<div class="im"><br>
> Is it important that the message is processed twice, or that it be<br>
> processed by two different consumers?<br>
<br>
</div>It's important that the message is processed/accepted by 2 nodes. Each<br>
nodes will be in its own machine hand have a limited number of<br>
specific message it can treat.<br>
<font color="#888888"><br>
- benoit<br>
</font></blockquote></div><br>