Ken Causey <ken at ineffable.com> wrote:

[snipped example message]

Excellent.  Let's go:

> To demonstrate my first method:
> 
> a body parts second body parts
> 
> returns #().  This method uses  MIMEDocument>>parts

Okay.  I don't know why.  I don't especially care about trying to make
this work.  MailMessage>>parts is better.


> 
> To demonstrate my second method:
> 
> a makeMultipart parts second makeMultipart parts first

Let's see.

First, the makeMultiparts are distracting.  They do nothing if the
message isn't multipart already, and if the message *isn't* multipart,
it will force them to be!  That is undesired.

But the main issue is that your message is actually structured like
this:

  a
     b  ("forwarded email with attachment)
     c  (a single-part "multipart" message)
       d  (an embedded message)
          e  ("original attachment")
          e  (test.txt)

Note that there is an extra layer here.  Since it is encoded as a
message, you'll have to decode it with "MailMessage from:" before
proceding further.  The following works to extract the attachment:


    forwardedMessage := MailMessage from:  a parts second parts first
body content.
    forwardedMessage parts second




> 
> at this point you can keep appending 'makeMultipart parts first' and you
> continue to get back the same exact result.  This method uses
> MailMessage>>parts.

In fact, the makeMultipart is more than distracting -- it is increasing
the nesting depth, even as the "parts first" is decreasing it! 


Overall, the difficulty is with the difference between a mail message
and its textual representation.  Okay I'm stopping now before I get a
headache.   :)

Lex