[rabbitmq-discuss] [announcement] amq-client 0.8.5 is released
michael.s.klishin at gmail.com
Fri Oct 28 16:40:15 BST 2011
Le 28/ott/2011 à 19:33, Matthias Radestock a écrit :
> Haven't you just introduced a server-side leak, assuming clients stay around for long enough? In which case the cure may be worse than the disease.
I understand this. Switching to some kind of default consumer later won't break any existing code so we almost certainly will do that. I am not sure what such consumer would do with messages that require explicit acknowledgements but I guess I can easily find that out in the Java client source and do the same thing.
Given that this scenario only happens when consumers are added and removed aggressively and explicit acknowledgements are used, to me it sounds like a reasonable trade-off that wins us time.
More information about the rabbitmq-discuss