On Mon, 29 Dec 2008 21:54:52 +0100, I wrote: >... I was surprised to see that when the recipient address >provided by Mailman is rewritten by Postfix' virtual_regexp, then the >recipient address that Postfix encodes in the envelope return path is >the rewritten address, rather than the original subscriber address that >Mailman knows.
I have just realized that there is another way to look at this, which may be a better argument for the semantics I would like: The problem occurs only because the sending server and the receiving server is the same; the recipient address is in a domain handled by the same postfix instance that Mailman uses to submit mail. If there were two independent postfix instances, this would not happen. In such a case, it seems to me that the result ought to be the same as if processing clearly related to the sending side, such as VERP address generation, happened before processing clearly clearly related to the receiving side, such as recipient address rewriting in virtual_maps. I.e., VERP belongs to "sending processing" and its result should therefore not depend on virtual_maps rewriting, which are part of the "receiving processing" and thus belongs logically "later"; it comes into effect in the same postfix instance only because the subscriber happens to be a local user. (But as I wrote earlier, I can live with the current semantics, and this will - probably - be my last attempt to convince you that the order ought to be different.)
