Re: [Evolution] Re: [HC Evolution] toolbar

[Jamie Zawinski <jwz jwz org>]

Tuomas Kuosmanen wrote:
> On Tue, Mar 28, 2000 at 01:18:42PM -0600, Fernando Romo wanted to say the following:
> > Dan:
> >
> > Think also in a "Redirect"  button. Maybe is a good feature implement this
> > function in Evolution. This take the message and forward to a third person.
> > The third person read the mail and appear to be send to this people by the
> > original sender and mark the message like "assign to". Very used in office
> > work. This Feature exists in Eudora 4.x
>
> Most mailers can do this. It is also called "bounce message".

Yes, and I believe that implementing it in a mailer intented for
non-guru users is a big mistake.  I intentionally didn't implement this
command in Netscape Mail, and I'm sure this was the right thing to do.

These two forwarding commands differ in the hint that the forwarder 
is giving the recipient about where replies should be directed. 
"Forwarding" makes it easy to send replies to the forwarder, and
"Resending" (the proper name for what you've called "Redirect", 
"Assign" and "Bounce") makes it easy to send replies to the original
sender.  

The problem is with the failure modes: when you have two forwarding
commands, people will often pick the wrong one.  The failure mode of
having used Forward when Resend was meant is that the reply goes to the
forwarder instead of the original sender.  The failure mode of
inappropriately having used Resend is that the reply goes to the
original sender.  This can be potentially disasterous, since it exposes
more information.

Person A sends a message to person B, who wants to forward it to person
C.  There are a number of ways this can be done: two legitimate ways,
and a whole bunch of shitty ways:

  1:  As a message from B to C with a message/rfc822 attachment:

        From: B
        To: C
        Subject: [Fwd: subject A]
        Content-Type: message/rfc822

        From: A
        To: B
        Subject: subject A

        Oink.

  2:  As a "resent" message from B to C, where the message appears
      to be coming from A:

        Resent-From: B
        Resent-To: C
        From: A
        To: B
        Subject: subject A

        Oink.

  3:  A variation of #2 that Eudora used for years (and still might)
      that is just totally gratuitous, nonstandard, and non-parsable:

        From: A (by way of B)
        To: C
        Subject: subject A

        Oink.

  4:  Worst of all, by just quoting:

        From: B
        To: C
        Subject: Re: subject A

        >From: A
        >To: B
        >Subject: subject A
        >
        >Oink.

The thing that you have to be very careful about with forwarded messages
is that the recipient realize who the message actually came from, and
that when they reply, that the reply is going to the person they expect
it to go to.  This is extremely important, because it's *very* easy for
people to embarass themselves by sending inappropriate replies, so your
mailer has to do what it can to help them.

#1 is the best forwarding method, because it is the clearest and most
versatile.  

  * First, it shows in the From and To fields who the actual sender and
    recipient of *this* message are, meaning that the failure mode for
    not paying attention and just hitting reply is a safe one: the reply
    goes to the forwarder, not the original sender (the person who has
    already corresponded with both B and C.)

  * Second, it can preserve the original message in its entirety,
    including Message-ID and routing info.

  * Third, by giving you the entire original message, it gives you a
    place to hang all possible reply commands: you can perform a gesture
    on the outer message to reply to the forwarder and/or forward
    recipients; or you can perform that gesture on the *inner* message
    to reply to the original sender and/or original recipients.   That
    is, it lets you reply to the appropriate message: the forwarded
    message, or the original message.

#2 is dangerous:

  * It encourages the recipient to reply to the original sender instead
    of the forwarder, and that exposes more information than the
    recipient might want to do.  E.g., replying to a customer when they
    meant to reply to a co-worker.  

  * It makes it far too easy for the recipient to fail to notice the
    "Resent-" headers (if they're lucky enough to have a mail reader
    that displays those headers at all) and think that A sent a message
    to C, when A actually sent a message to B.  A might not even know C,
    and when C replies to A, confusion will result, especially if C
    thought that A sent the message to C.

  * It makes it harder to reply to the forwarder and/or the other
    recipients of the forwarded message.

#3 is just not even worth talking about.  It makes it impossible to do
anything useful with the message, and I have no idea why they did things
this way for so long.  (They did this from 1992 through 1996, possibly
longer.)

#4 is what people use when their mailers don't support sensible
forwarding commands.  It's almost as bad as #3 (not quite as bad because
at least the same information is present as in #1, even though you have
to piece it together by hand with cut and paste.)

So.  It is my humble but correct opinion that your mailer should
implement #1 as the one and only forwarding style; and that your mailer
should make it easy to reply to attached messages.  That solves both
sides of the problem, and does so in a well-structured and
standard-compliant way.

"Eudora does it" or "Exchange does it" are not good justifications for
implementing forwarding style #2.  It's a bad idea, and you're not doing
your customers any favors by implementing it.  

Neither is "the customers asked for it" a good reason to implement it:
customers *don't* know best, and they ask for all kinds of things that
would hurt them if they got it.  (For example, customers asked for the
ability to post HTML messages to USENET, and Netscape complied.  Do you
still think the customer is always right?)

-- 
Jamie Zawinski
jwz jwz org             http://www.jwz.org/
jwz dnalounge com       http://www.dnalounge.com/