Re: [Dcrup] I-D Action: draft-ietf-dcrup-dkim-usage-00.txt

[Jim Fenton <[e-mail]>]

On 6/5/17 1:18 PM, Murray S. Kucherawy wrote:
> On Mon, Jun 5, 2017 at 9:53 AM, Scott Kitterman <[e-mail]
> <mailto:[e-mail]>> wrote:
>
>     I hear you.  I expect this to be the most contentious part of the
>     draft.
>
>     Here's my counter argument:
>
>     The first DKIM RFC (RFC 4871), published in 2007 said:
>
>     > Signers MUST implement and SHOULD sign using rsa-sha256
>
>     I believe that the only reason rsa-sha1 was included at all was to
>     make
>     transition from domainkeys easier (see RFC 4870).  That's also (as I
>     understand it where the 512 bit minimum key size came from).
>
>
> As I recall, there were also implementations of DKIM made during its
> development (even before its IETF time) that supported both, and
> defaulted to rsa-sha1 because support for SHA256 in OpenSSL was new
> and not universally deployed.  Some of them were still running in the
> wild, and SHA1 wasn't fully deprecated, so the choice was made to be
> inclusive while encouraging use of the newer stuff as much as possible.

That matches my recollection. There was an effort to maintain DomainKeys
compatibility for selector (DNS TXT) records but the choice of hash
doesn't affect them. The signatures were different anyway.
>  
>
>     As a working group document editor, I'll change it however the
>     group wants (of
>     course), but I think we should either kill rsa-sha1 entirely in
>     this document
>     or leave it out entirely and let one of the follow-on documents
>     add a new
>     algorithm and remove rsa-sha1.  Preferably a clean kill or,
>     failing that, not
>     at all is what I think we should do.
>
>
> I would omit it from the updated version entirely, and mark it
> "obsolete" in the registry.

Concur.

-Jim