Re: email standards (was: Re: facilitators at [e-mail])

[Scott Kitterman <[e-mail]>]

On Tuesday, September 23, 2014 16:48:29 John C Klensin wrote:
> --On Tuesday, September 23, 2014 16:08 -0400 Phillip
> 
> Hallam-Baker <[e-mail]> wrote:
> >> Surely PHB isn't saying that SMTP and the email format docs
> >> are incompatible?  That would be a nonsensical assertion,
> >> since they are separate layers (the one is used to transport
> >> the other).  Perhaps there are two different email standards
> >> that perform the same functions but are incompatible?
> >> Perhaps S/MIME and PGP?  Or perhaps two different security
> >> related email specs?
> > 
> > I meant two secure email standards. Empirically we have two
> > right now, S/MIME and PGP.
> > 
> > Since I was talking about security, I thought it was obvious
> > from the context.
> 
> Nothing about your note made that clear -- it didn't mention
> security generally and you said "email standards".  Be that as
> it may, I think you are overlooking a key aspect of the PGP
> versus S/MIME problem.   Suppose we actually did have two sets
> of email standards, one using SMTP transport with 822-style
> "field-name: value-string" headers (as we have today) and other
> other of which used SMTP (to avoid making this completely
> unrealistic) with ASN.1-like coded X.400-like inner envelope
> header structure.  There would certainly be a reasonable
> complaint that we had specified two different ways to do the
> same thing with only subtle differences in capabilities between
> them.
> 
> But it seems to me that S/MIME and PGP represent two
> fundamentally different trust models.  The first is based on a
> certificate hierarchy model, one that would have very good
> international scaling properties had we actually figured out how
> to make a global single-purpose PKI work and be trusted.  Worse,
> absent that type of PKI, it was very hard to think about how to
> bootstrap the system, at least without pushing decisions about
> which certification authorities to trust back to end users who
> had absolutely no basis on which to make those choices.  The
> second is based on a web of trust arrangement that most of us
> knew at the time wouldn't scale well internationally nor be
> usable among parties who didn't have at least a second, or
> possibly third, "degree" of connection but that was far easier
> to bootstrap than something that assumed a global PKI.
> 
> Now it is certainly possible to imagine a message format that
> would have more commonalities than we ended up with.  We
> actually had standards-track specifications for such a format,
> in the form of RFC 1421ff and the earlier RFC1113ff.   I think
> it is reasonable to summarize PEM by saying it went nowhere
> except that we might have learned a bit from it in building
> S/MIME and/or OpenPGP.
> 
> So, we are now at a point at which neither OpenPGP nor S/MIME
> has achieved wide adoption and use.  We have learned such things
> we (at least some of us) didn't anticipate.  In S/MIME's case,
> that notably includes issues of trust in CAs and the
> effectively-dictatorial (or oligarchic) authority of browser
> vendors to determine CA usability.  In OpenPGP's case, we have
> demonstrated some of the scaling and key management issues that
> some people anticipated all along.
> 
> You seem to believe that more commonality of formats would have
> left us in better shape today.  Because I think the problem is
> the irreconcilable difference in trust model and relationships,
> I believe it would have made almost no difference at all (even
> if it were a good idea).  You could be right but, if you want to
> make that case, please try to do so in a way that the rest of us
> can understand rather than, e.g., making broad assertions about
> causes and implications of the IETF's failure to generate a
> single standard for secure/encrypted email or email more
> generally.

+1.

I use both OpenPGP and S/MIME on a regular basis and in no case where I use 
one would the other be suitable primarily because of the differences in trust 
models you describe.  While they both sign/encrypt email their use cases are 
disjoint in my experience.

Scott K