(Reposted to the list as per Murray's suggestion.)
Please do feel free to review what's there and comment, and also submit
suggestions for other cases that might be of interest to record for
future implementers.
A couple of things:
- an example of non-valid header I have seen "in the wild" (in legitimate email) is a non-encoded non-ASCII header (something Russian, I believe). I'm not quite sure how this can be abused though, but it's probably worth mentioning. I know at least one spam filter that does (or did; I think they changed it) block such messages outright as non-valid email.
- it's not mentioned in the document, but in the related discussion on the DKIM-list, "actions" to be taken by MUAs were implicitly or explicitly mentioned. I just wanted to say I don't think a MUA _should_ do anything but, perhaps, render a message in a specific way. Many people have filters built in their MUAs, but many others haven't.
- it would be great if the document could somehow say it is okay for a filter to discard/deny/drop certain kinds of messages. (Multiple From-headers, for example.) I know many spam filters don't dare to do this, as there is always the risk of false positives and the need to be liberal. A document like this could give them some kind of official 'permission' to block these messages which would, hopefully, encourage both legitimate senders and spammers to not send them.