ietf-smime
[Top] [All Lists]

Re: Is 7bit restriction to signature practical?

1997-09-10 22:40:22
Some sons of multipart/signed(aka RFC1847) including PGP/MIME and
S/MIME restricts an object to be signed to 7bit. From the UA
programmer's point of view, I wonder if this restriction is practical.

Let me give you an example.

In Europa, we can find many 8bit messages as follows:

      Subject:
      From:
      To:
      MIME-Version: 1.0
      Content-Type: text/plain; charset=iso-8859-1
      Content-Transfer-Encoding: 8bit
        
      ISO-8859-1 text comes here.

When I try to forward this message with my signature, how should I do?

Simple: You change the encoding from 8bit to quoted-printable.

I mean that I want to create the following message.

      Subject:
      From:
      To:
      MIME-Version: 1.0
      Content-Type: Multipart/Mixed; boundary=foo

      --foo

      An explanation on the following forwarded message.

      --foo
      Content-Type: Multipart/Signed; boundary=bar;
              protocol="yyy"; micalg=zzz
      Content-Transfer-Encoding: xxx

      --bar
      Content-Type: Message/Rfc822
      Content-Transfer-Encoding: xxx

      Subject:
      From:
      To:
      MIME-Version: 1.0
      Content-Type: text/plain; charset=iso-8859-1
      Content-Transfer-Encoding: 8bit
        
      ISO-8859-1 text comes here.

      --bar
      Content-Type: yyy

      Signature comes here.

      --bar--
        
      --foo--

Since 7bit restriction is required, I think the forwarded message must
be first analyzed with MIME syntax then be converted to 7bit.

Correct.

Is this practical? How about forwarding a more complex MIME message
which is in 8bit?

Well, let's see. In order to conform to the requirements of SMTP an agent must
be prepared to do this whenever it encounters an SMTP server that doesn't
support the 8bitMIME extension. (In fact we're starting to see SMTP server that
refuse to accept unlabelled 8bit, because so many problems attach to 8bit that
doesn't have proper charset labelling.) So what you're asking isn't whether or
not user agents can support multipart/signed, you're asking whether or not user
agents can support MIME. And the evidence is pretty darned strong that they
can.

Now, if you want examples of actual implementations, you'll find that the
code necessaary to do this is part of modern versions of sendmail. I think
the fact that this functionality is part of one of the most, if not the
most, popular email packages around pretty much demonstrates that not
supporting it is inexcusable.

I guess many implementations just include messages when forwarded.

My various agents certainly do not. Again, this has nothing to do with
multipart/signed per se; it is just a basic part of MIME functionality.

                                Ned

<Prev in Thread] Current Thread [Next in Thread>