Fully agree with you. S/MIME as-is is not usable in the wild.
PGP is better, but I doubt its usable by an entire organisation
or set of home-users for most mail messages.

Let's posit an end goal where a large proportion of email ends
up with ciphertext bodies at least, but typically without a high
level of assurance that the right public key was used.

My questions:

1) Would we like to get there?
2) If we would, should we start from a) S/MIME or b) PGP or
   c) yet another mail-security clean slate?
3) If we got as far as having IETF consensus on something
   we think could do the above (i.e. we completed the RFCs
   for (2)), do we think there's any real chance that that
   could be deployed and end up widely used?

FWIW, my answers would be 1) yes, 2) (c), reluctantly, since
it'd only be worthwhile if it covered headers, and 3) "I very
much doubt it unfortunately."

I'd be surprised (but pleased) if we had a rough consensus on
these questions. And even more surprised and pleased if that
consensus indicated that there's an obvious thing that we
should start doing in the near future.

Sorry to be pessimistic, and I'd really love to be wrong, but
I can't see a realistic way to get e2e email confidentiality
out there as of now.

If the top-tier email providers got their act together on
something that didn't allow 'em to scan mail content (without
launching a MITM) then that'd change my pessimism. I guess
there is some chance of that, but I suspect that the
initiative for such a service-provider driven effort would
start outside the IETF (same as DKIM/DMARC).

S.

PS: Clarifications: SMTP/TLS with DANE is still worth doing
and is being done. S/MIME and PGP are quite usable within a
community or even large enterprise, but not on the big-I
Internet. e2e mail message signing is maybe more easily
doable but is uselessly boring IMO and not relevant for
perpass, so that's not a goal at all. And by e2e confid.
I mean that in almost all cases the private decryption
keys are only really usable/available to an MUA on the
user's machine and almost never on the MTA/MS.




On 09/07/2013 02:39 PM, Yaron Sheffer wrote:
> Hi,
> 
> I have wanted to get my company on S/MIME for a while, and the recent
> noise was the final motivator I needed. We are a small company doing
> security, however (like anywhere else) not everybody can be considered a
> security "expert".
> 
> So Outlook and Thunderbird have good support for S/MIME. This is a good
> starting point, right? Personally I am using Thunderbird running on
> Linux, which has very convenient S/MIME support. I had actually used it
> in the past.
> 
> Below I will show that in today's market you simply cannot use S/MIME,
> because of a combination of bad security practices, silly web-site
> design, lousy CA support on Linux and probably a few more factors.
> 
>  * Started with the free options. The Web is full with tutorials on how
>    to install the free Comodo email cert in your mail client. It turns
>    out, with InstantSSL (Comodo) you cannot register twice with same
>    email address (e.g. if the cert is lost for some reason or you just
>    want to use two different machine without shuttling private keys
>    around). The same is true for StartSSL.
>  * Next tried Symantec: this is $22 per year, the UI is not very good
>    (says cert is installed but then has a button to install cert). TB
>    says the certificate could not be validated "for unknown reasons". I
>    guess there is no valid certificate chain. Well, Symantec doesn't
>    appear in either the Chromium/Linux or Firefox/Linux cert stores.
>  * GlobalSign: EUR 12 for 1 yr, 29 for 3 yrs. Not too bad. So you go
>    into their wizard. The default is that the private key is generated
>    by the CA! Which means this product is not (securely) usable for
>    multiple users in an organization. Most of them will probably leak
>    their private key.
>  * CACert: Free and open source. Probably still struggling (the server
>    is extremely slow). Surprisingly, the CAcert root CA is known by
>    Chromium/Linux but not by TB/Linux (stock Thunderbird on Ubuntu 12.04).
>  * Entrust: pricing is only for US, UK and Canada. Other customers are
>    referred to a small number of resellers (none for my geography).
>    They still let you order the cert though. And then surprise! The $20
>    price that appears on the "Buy Now" page turns into $30 when you
>    complete filling the form.
> 
> This covers all I could find on the first 4 Google search pages for
> "email certificates". I will try again in a year or two.
> 
> 
> Thanks,
> 
>     Yaron
> 
> 
> 
> _______________________________________________
> perpass mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/perpass
> 
> 
_______________________________________________
perpass mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/perpass

Reply via email to