What's this with formail -r?

Here's what the relevant part of the formail source code looks like:

/*
 *	sender determination fields in order of importance/reliability
 *	reply-address determination fields (wrepl specifies the weight
 *	for regular replies, wtrepl specifies the weight for trusted users)
 *
 *	I bet this is the first time you see a bar graph in C-source-code :-)
 */
static const struct {const char*head;int len,wrepl,wtrepl;}sest[]=
{ sslbar(replyto	,"******"	,"********"	),
  sslbar(Fromm		,"*"		,"*******"	),
  sslbar(retreceiptto	,"********"	,"*****"	),
  sslbar(sender		,"*****"	,"******"	),
  sslbar(res_replyto	,"***********"	,"***********"	),
  sslbar(res_from	,"***foo***"	,"***bar****"	),
  sslbar(res_sender	,"**********"	,"*********"	),
  sslbar(errorsto	,"*******"	,"****"		),
  sslbar(path		,"**"		,"*"		),
  sslbar(returnpath	,"***"		,"***"		),
  sslbar(From_		,"****"		,"**"		)
};

In so many words, this is how it comes out:

formail -r

Resent-Reply-To:
Resent-Sender:
Resent-From:
Return-Receipt-To:
Errors-To:
Reply-To:
Sender:
From_
Return-Path:
Path:
From:

formail -rt

Resent-Reply-To:
Resent-From:
Resent-Sender:
Reply-To:
From:
Sender:
Return-Receipt-To:
Errors-To:
Return-Path:
From_
Path:

In particular, notice how formail -r will prefer the From_ pseudoheader over the real From: if nothing much else is present. Also note how Errors-To: (which is deprecated) is preferred over Reply-To:.

In short, the "trusted" meaning of the -t flag is largely bogus. Briefly, -r would be most appropriate for returning error messages and bounces, whereas -rt should be preferred for "real" replies.

(Even so, some of the priorities here are somewhat strange.)

$Id: formail.prep,v 1.3 1999/08/23 07:25:43 era Exp $