Message-Id: <199510040611.CAA25974@thud.cs.utk.edu>
From: Keith Moore <moore@CS.UTK.EDU>
To: drums@CS.UTK.EDU
cc: moore@CS.UTK.EDU
Subject: Re: What's the sender header for? 
In-reply-to: Your message of "Wed, 04 Oct 1995 00:29:13 CDT."
             <9510040529.AA14571@dogie.macc.wisc.edu> 
Date: Wed, 04 Oct 1995 02:11:04 -0400
Sender: moore@CS.UTK.EDU

> I think that CC: headers are supposed to convey the notion
> "for you information", which means to me that such folks are
> not expected to participate in any discussions.  Why would
> they expect to see replies?  Even more so for Bcc: recipients;
> no one is supposed to know they're eavesdropping.

Whether CC recipients expect to see replies isn't defined by
the standard.  It's up to the author (who can specify his
preference by using reply-to), and the recipient doing the
reply (who can send it to anyone he wants to).

Nobody expects Bcc recipients to get replies.  The only purposes
of the Bcc header appear to be (a) to allow the sender to specify
Bcc recipients when the message is submitted (if the envelope
is derived from the header), and (b) to allow a Bcc'ed recipient 
to see why he got the message.  (Whether the Bcc'ed recipient
can see the addresses of *other* Bcc'ed recipients seems to 
depend on the implementation.)

> I'm not arguing for anything here as much as I'm asking questions.
> If we have To: CC: and Bcc: headers for recipients, they must
> have different meanings, otherwise why have them?  

The difference between To and CC may not be preciesly defined,
but they're not precisely defined for paper mail either, and 
they're still useful.   A somewhat precise answer as to when 
to use To and when to use CC might be found in a "style" or 
"etiquette" or "protocol" guide for human communications, but
exactly what To and CC mean are probably context-dependent.

> I realize that you often want a reply to go to "everyone that
> saw this message" (like this one).  But that doesn't  necessarily
> mean to me that a UA should ferret out all the To: CC:, etc.
> addresses and send replies there; it might mean that the message
> was improperly composed.

Yes, but you really don't want the recipient's UA trying to
second-guess the sender!  In general, we try to avoid defining
the user interface, so we don't tell you who to reply to.
All we say is (a) in principle, you can reply to whomever
you want to, and (b) the author's preference about where to 
reply can be found in the Reply-To header. 

> >And more to the point, where does MAIL
> >FROM come from?  Do you generate it from the Sender field?)
> 
> Any reason it doesn't come from the same place it would come from
> if you were just creating and sending a message?  The only case I
> can think of when you would need to generate it from RFC822 headers
> is if the message is being gatewayed from a non-SMTP envirinment,
> in which case we're going to say get it from the Return-Path: header,
> right?

Lots of people compose messages with RFC 822 and then hand it to
sendmail or some other MTA with a special option that means 
"submit this message".  The MTA then groks the headers and builds
an envelope from them.

In general, when a message is gatewayed from a non-SMTP environment
you DON'T want them generating an envelope from the headers.

(Sigh) There's one FidoNet (I think) site that got its mail from
a UUCP link, that kept bouncing messages back to me (as maintainer 
of the info-mime list) because it didn't accept responsibility to 
deliver the message to info-mime@cs.utk.edu, which appears in the 
To or Cc header of every message.   It took me about 4 or 5 messages 
to get its postmaster to understand that you don't deliver to the
addresses in the message header, and where to look for the envelope
recipient list.

Keith