Open Stack

On Mon, Jul 15, 2013 at 4:28 PM, Russell Bryant <rbryant at redhat.com> wrote:

> On 07/15/2013 02:36 PM, Doug Hellmann wrote:
> > The namespace relates to the API implementation inside the receiver. The
> > way it currently works is the receiver subscribes to messages on a
> > topic/exchange pair to have AMQP route messages to it, and then it looks
> > inside the message for further dispatch to an object that knows about
> > that API. That lets the nova API implementation be split up among
> > different objects, for example. I'm not sure why it evolved that way,
> > instead of using separate topics and having the messaging layer do all
> > of the routing. Maybe we should take another look at that part of the
> > new API design.
>
> In retrospect, yes, a separate topic would have worked.  The namespace
> was very convenient for the current nova implementation, but that
> doesn't mean it was the best design.  The code that sets up which topics
> to consume from is very generic and applies to *all* services.  So,
> instead of reworking this to let it be different per-service, I did the
> namespace thing, which worked without having to change any other nova code.
>

That's a completely understandable pragmatic solution. :-)

Doug


>
> --
> Russell Bryant
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130715/ba334038/attachment.html>