[Fwd: Re: [ietf-dkim] canonicalized null body and dkim]

[Eric Allman]

--On December 15, 2006 10:17:16 AM -0800 Mark Delany 
<markd+dkim at yahoo-inc.com> wrote:

> FWIW, this problem was similarly discovered in DK. The early text
> read:
>
>
> -01
>     o All trailing empty lines are ignored. An empty line is a line
> of
>        zero length after removal of the local line terminator. The
>        empty line that separates the header from the body is a to be
>        included in this process.
>
>
> and the later text read:
>
> -06
>      o All trailing empty lines are ignored. An empty line is a
> line of
>        zero length after removal of the local line terminator.
>
>        If the body consists entirely of empty lines, then the
>        header/body line is similarly ignored.
>
>
> In short, if the last empty line of the email is the header/body
> separator, then it should not be fed into the canonicalization.
>
>
> The "simple" in DKIM, as I understand it, is merely re-codifying the
> same function.

I *think* there are three cases here:

(1) Some body with l=0.  I think we're agreed that this should result 
in an empty input into the hash algorithm.

(2) Header, CRLF, no body; that is, the input to a DATA would be:

        Header: foobar<CRLF>
        <CRLF>
        <CRLF>.<CRLF>

The last line of the example is the separator, so the initial CRLF 
doesn't count.

(3) Header, no CRLF, no body:

        Header: foobar<CRLF>
        <CRLF>.<CRLF>

It seems to me that all three of these should match after 
canonicalization.  I'm not sure the -07 draft is explicit enough to 
ensure this.

In short, I'm not sure we have guarantied DK compatibility based on 
the current wording, regardless of intent.

eric