Re: [Nmh-workers] Dealing with missing From: header during send.

[Tethys]

Ken Hornstein writes:

>I understand his point, but my point is that while in theory it may
>be possible to ask Google for some new header, in practice I strongly
>think any request like that would just get ignored.  I'd be happy to
>be proven wrong, though.  So we have to work with the tools that
>we have today.

Agreed, to a point. It's not the tools that you have to work with
that are the problem but the data. You need to find out the recipient's
address before you can reply using the same address. Looking at the
envelope is the best way to get that. In a "normal" mail message, that
information simply isn't present, so we need some way of approximating
it. Once we have an address, putting it into the composed message is
trivial.

Google might actually be receptive to adding an appropriate header.
Who knows? But even if they are, there will be plenty of others that
aren't, and we need a solution to deal with those cases. How about using
X-Envelope-To if it's present (I didn't make the header name up, it
was already in widespread use with that name[1]). If it's not, you could
possibly attempt to parse the Received headers, or maybe look for the
last Delivered-To header if present. Neither are ideal, but lacking the
envelope addressee, they'll probably give you a reasonable best guess[2].
Or at least a better guess than looking at the To/Cc headers. Maybe
allowing a user to hardcode an address to use with regexps to select
from a list would be good enough for most. Personally, I use over 500
addresses, so it wouldn't work well for me without a fair amount of
work :-)

>>On the other hand, Tethys' X-Envelope-From is just Return-Path and
>>should be there already.
>
>Yup, that's true.

Yep. I added it for symmetry because I was adding X-Envelope-To at
the same time, without thinking about whether the information was
already there in another header.

Tet

[1] And any other similar headers in common use that do the same thing.

[2] But parsing Received headers correctly is *hard*.