Re: IESG Review: draft-ietf-fax-esmtp-conneg-10.txt

At 18:12 10/07/04 +0700, Dave Crocker wrote:
>I propose to add to:
>
>   1. Content Negotiation section:
>
>   Content conversion will invalidate a message's signature and will not
>   be possible to perform with encrypted content. Therefore, permission
>   to convert SHOULD NOT normally be given with signed or encrypted
>   messages.

Dave, I recognize that this reflects, in part, a comment that I made.  But 
on reflection I find that I'm uncomfortable with the implication that 
content negotiation and security are mutually exclusive options.

In the case of encryption, as you note elsewhere, the conversion can be 
performed by an intermediary with access to the appropriate keys.

In the case of signature, I can imagine the possibility of a trusted 
intermediary that performs a conversion and then signs it as being a 
conversion of some original document.  In such circumstances, maybe the 
original signed document should also be delivered.

I think these are significant issues to explore with respect to content 
negotiation and security, but also think there's a reasonable concern about 
how much detail is required.  I'm thinking that the text should cover the 
following points (and avoid the SHOULD NOT used above):

1. If a document-converting intermediary fails to ensure that the resulting 
converted content is similarly protected, then the benefits of the 
originally applied security are lost.

[Note:  this comment does not exclude the possibility that the conversion 
and onward transmission occurs without added on a trusted network.]

2. Conversion of an encrypted document requires a (presumably trusted) 
intermediary to have access to the decryption key.  Such intermediaries 
should take steps to ensure the resulting concerted content is similarly 
protected (e.g. by re-encryption).

3. Conversion of a signed document results in a document that is not 
covered by the original signature.  An intermediary that performs 
conversion should assure some level of integrity is applied to the result 
of the conversion (e.g. by itself signing the converted document together 
with a hash of the original, and possibly including a copy of the original 
signed document).

4. A full discussion of mechanisms for applying security to converted 
documents is beyond the scope of this specification.

[But maybe a mini-survey, drawn from the comments above?]

Maybe it should also be stated that, in any case, the intermediary MUST at 
least pass information to the ultimate recipient indicating that the sender 
applied encryption and/or signature to the original content?

#g


------------
Graham Klyne
For email:
http://www.ninebynine.org/#Contact