[rabbitmq-discuss] request for help!

[Rafael Schloming]

Gordon Sim wrote:
> On 05/06/2010 05:12 PM, Alexis Richardson wrote:
>> On Thu, May 6, 2010 at 5:07 PM, Gordon Sim<gsim at redhat.com>  wrote:
>>> I would expect an implementation to define policy around where links
>>> with these two types of distribution mode can be used. E.g. a pub-sub
>>> topic might only allow non-destructive links were valid, a plain queue
>>> might only allow destructive links and a browsable queue might allow
>>> both (with what I think are then obvious if non-deterministic
>>> implications for concurrent links with different modes).
>> Is there any use case where 'destructive' or 'non destructive'
>> behaviour should be a property of the LINK rather than the NODE?
> 
> Browsing a queue (you would want to allow consumers and browsers, though 
> most likely at different times).

One point of clarification, the destructive/non-destructive property is 
not actually associated with a link, it's actually part of the source 
specification. There is no distinction between destructive and 
non-destructive links from the perspective of the transport. It makes no 
difference to the transport whether the result of an ack is to advance a 
cursor or to dequeue a message.

The distribution-mode field of the source specification simply permits 
the application to indicate how the link is/should be treated by the 
node, in other words destructive/non-destructive is not a "property of 
the link", but rather a classification of the outgoing links from a 
certain kind of node, and this classification has no impact on how links 
function, only on how the node behaves with respect to the two different 
classes of outgoing links.

--Rafael