Hi I have looked through and tried patches 1 and 2 and it looks OK to me. So at least a sort of +1 on them combined (I didn't really look at them separately as they seem to make more sense as 1 lump). I agree with Jani about both the date option, and deduplicate possibility, and we may also want to add an option to use --output=sender. However, I think anything like that could come as a later patch. For the moment the only thing I think we need to consider is the variable notmuch-address-command. Do we want to have something other than nil to mean use the elisp completion? In particular we could have something which can be extended to specify the options above. Additionally, aesthetically, I like the idea that setting this variable to nil disables address completion. Best wishes Mark