> Spool is just a specialised repository which is FIFO and normally empty. > Yeah, we use JMS queues and our MailListeners are responsible for dealing with them via name injection + jndi lookup.
I've often though of JMS for Spools, but again I think it should be an implemenation detail, in the same way that JMS queue implementations can be backed by rdbms, or in memory or technology X. Put it this way, I think the sucess criteria would be that the RI could be written which did not require rdbms or JMS or anything that might presume "vendor"s' implementation decisions.
> Yeah, but consider this, map POP3 IP listener addresses to domains and > you can have two "andy" accounts on the same instance, but they will > be discrete. > This is kind of an oddball thing. It is not a feature that is common among mail servers that I've seen (most of my research has been in postfix, JAMES, Exchange and dovecot).
Its a recurring theme in James users requests for enhancements, has been since the year dot.
Basically you have a concatenated principal (in security speak) of account + IP. I don't think the Mailet API should require servers to have this feature
Completely agree, with my two hats (Mailet & James) swapping like crazy I have to say that I want the Mailet API to be capable of supporting this requirement, but not for it to have any direct refrence. <snip, illustration with IP addresses in it/>
The lovely thing is mailets that are unaware of this do not have to parse it just "nope not the same account".
And so it should be... d.
