Re: [ietf-dkim] Re: dkim service

On Oct 13, 2005, at 4:20 PM, Jim Fenton wrote:

> Dave Crocker wrote:
>
>
> > > Right.  The idea, as I put it once, was that "If you break it, you
> > > bought it."  Put less colloquially, if a mailing list that knows  
> > > it's
> > > going to mangle the message receives a DKIM-signed message, it  
> > > should
> >
> > this was/is a particularly important view, since it relieves the  
> > signing effort of quite a bit of responsibility that would  
> > otherwise require an impossible effort to withstand arbitrary  
> > modification.
> > a derivative bit that emerged from this was the counter-intuitive  
> > possibility that a mailing list capable of doing dkim signing  
> > could choose not to, for a message that is already signed.  that  
> > is, it could knowingly preserve the existing signature.
> This relates to one of the motivations for multiple signatures.  If  
> you have a non-mangling mailing list, you might want to preserve  
> the original signature, because it's still valid and some people  
> might want to base a decision on that.  They (or others) might want  
> to know for sure that it came from the list, because they want to  
> make sure that they read all messages on the list.  A WG chair  
> might have that concern, for example.
Agreed, but how would you ensure the sequence of the keys?  The idea  
that Phillip expressed of self sequencing leaves this method open to  
strange exploits. Even by including a finger-print of the prior  
signature, there is room for games.
There could be a rule that signatures are removed beyond a point of  
discontinuity when deletion are obvious.  This would tend to retain  
the originating signature.  Of course one could lie about there being  
a signature missing, where this signature would be removed when being  
resigned?  Take the case where only a signature with the count of 1,  
which means the initial signature was removed?
I don't see an obvious way to prevent people from playing games with  
multiple signatures unless they are all treated as equivalent, but  
this adds a fair amount of overhead processing a stack of  
signatures.  Could there be a primary and a secondary signature  
scheme?  The Secondary signature would be the one removed when needed  
to resign a message.  This provides a first and last signer.  When  
there is only a secondary, the signer wants to be dropped at the  
first opportunity which should mean this message is refused.
-Doug
_______________________________________________
ietf-dkim mailing list
http://dkim.org