Jean-Marc Desperrier wrote:
Frank Hecker wrote:
"the CA must not knowingly issue certs to entities who do not own the
associated domains". (The language also has to address the question
of agents who are authorized to get certs on behalf of the someone else;
Which as Duane points in the real world of SSL certificate is an
excessively common occurence. I propose :
"the CA must issue certs only to an entity that have received
authorization from the associated domains owner"
* For object signing certs the requirement should be something like
"the CA must take reasonable measures to verify the identity of the
entity associated with the certificate"
Well I don't see why we don't want that for the other cases too ?
Email is the clearest example. You and I are exchanging this
email. Yet we have not identified each other. Let's say we
want to use S/MIME so we can exchange encrypted email
because we don't want our sneaky ISPs to read our mail.
To use S/MIME we need CA-signed certs. To get CA-signed
certs, we might need to identify ourselves. Why? What has
my identity got to do with wanting to share encrypted email
with you?
(As a matter of complete practicality, people don't think in
terms of identifying people in normal real life. Most of the
people I share PGP email with I never 'identify' in any formal
sense. In fact, I don't think I've ever identified anyone for
any email usage ever at all, and I've done quite a lot of
serious hard-dosh business over PGP.)
For code signing it's the same thing. I release a lot of code.
I'd like to release it signed (well, maybe). But I don't want
to have to 'identify' me, and the people who want to use
the code have no need to 'identify' me. They would rather
just be sure the code comes from me, which is a completely
different thing than knowing who I am.
Other people might have other needs. If I may drift into
crypto-politics, there are use cases in for example crypto
code - not so many these days thankfully - where identifying
the signer of code could be quite traumatic in the wrong
places & times.
What I think we need for object signing certs is not to tie X-owned
object signing certs to X-controlled domains, I don't think it makes
too much sense, but an insurance that the CA will process external
report that the software is acting badly, and accept to revoke it
based on that input.
If you ask for insurance from the CAs, I suspect they will
stop selling certs. Nobody can deal with the liability
aspects of bad software; especially in today's uncertain
world where security people like Bruce Schneier are
actively suggesting there be legislation on software
liability for producers.
iang
--
News and views on what matters in finance+crypto:
http://financialcryptography.com/
_______________________________________________
mozilla-crypto mailing list
[email protected]
http://mail.mozilla.org/listinfo/mozilla-crypto
