[rabbitmq-discuss] Help defining appropriate queue model

[Michael Bridgen]

Fabio,

>     Right, at the minute the last values are transient. It probably
>     wouldn't take much coding to make it save things to disk for durable
>     exchanges.
>
>
> OK, that figures... :) I know nothing of Erlang, but I'll take an
> overall look and try to come up with an acceptable solution for my case.
> Would that be a feature useful enough that you'd consider including in
> your code?

Sure!  Here's a sketch: use two mnesia tables, one ram only, for 
non-durable exchanges, and one on disk, for durable exchanges.  In a 
single node this will be relatively easy; if you use a cluster, you'll 
need to find out which nodes have on-disk state.


Michael