Re: [ietf-smtp] Fwd: New Version Notification for draft-crocker-email-deliveredto-00.txt

[Dave Crocker <[e-mail]>]

On 2/5/2021 2:56 AM, Alessandro Vesely wrote:
> 1)  In the sentence:
> 
>     The Delivered-To: header field is added at the time of delivery, when
>     responsibility for a message transitions from the Mail Handling
>     (Transport) Service to an agent acting on behalf of the specified
>     recipient address.
> 
> (Please substitute /Mail Handling/Message Handling/, so that it matches 
> RFC 5598.  Perhaps substituting /Transport/SMTP/ might improve 
> readability.)

oops.  good catch!


> That sentence implies there is no Delivered-To: if the message is copied 
> locally without going through MHS. 

I would claim that, in architectural terms, whatever code is doing this 
local processing, figuring out where to place the message, is part of an 
MHS.


  For example, some MDAs accept either
> email addresses (local or remote) or local mailbox paths to deliver a 
> copy.  If one specifies delivery by means of local paths, no further 
> Delivered-To: is added (and neither could be, since the agent doesn't 
> know the address in that case).  Is that worth noting?
> 
> BTW, you don't use the term Message Delivery Agent (MDA).  Is it by 
> purpose?

It was intentional.  Note that the text is MHS and not MTA.

One of the more interesting bits from the development of RFC 5598, was 
some folk putting forward the idea that the MSA and MDA each has a dual 
personality.  Part MUA; part MHS.  So the boundary for the MHS is 
/inside/ each of those architectural components.

See Figure 5 in RFC 5598, which makes this explicit.


> 2) ABNF
> 
>     "Delivered-To:" FWS Mailbox CRLF
> 
>     Note:    The field records only a single address, for one recipient.
> 
> Would it be worth to note that Mailbox (capital M) comes from [SMTP], 
> not [Mail-Fmt]?

A perfectly reasonable question, which also highlights the basic problem 
with defining shared constructs in multiple places.

But to respond to the question:

https://tools.ietf.org/html/rfc5321#section-2.3.11:

     Mailbox        = Local-part "@" ( Domain / address-literal )

https://tools.ietf.org/html/rfc5322, Section 3.4:

     mailbox         =   name-addr / addr-spec

Since the string comes from a RCPT TO, I think the ABNF details need to 
be for /only/ and address (no name) and so it's either an RFC 5321 
<mailbox> or it's an RFC 5322 <addr-spec>.  But since the source of the 
string is 5321, I'll choose the former, unless there's a strong lobby 
for the latter.  (Also, I think the existing use of the capital M 
signals that 5321's version was intended...)


> 3) IDNs:
> 
> When I write a message to user@foà.it (fake user, real domain) the MDA 
> writes
> Delivered-To: [e-mail].  I see no reason for doing so.  The 
> message is SMTPUTF8 anyway because of the To: user@foà.it header field, 
> which is maintained.  Are there reasons to mangle the domain name?

I'd guess that has more to do with software implementation that document 
specification, but if there's a change to this spec you have in mind, 
please offer it.


>> The SMTP mailing list is cited in the draft as the discussion venue.
> 
> 
> Didn't find that citation.

Section 2.


d/


-- 
Dave Crocker
Brandenburg InternetWorking
bbiw.net