Duane wrote:
Frank Hecker wrote:
Probably relatively small. I suspect that most usage of IMAP/SMTP over
SSL is by small sites like mine that are cost-conscious and thus aren't
using commercial CAs. People doing IMAP/SMTP over internal networks
(e.g., your typical corporation) don't necessarily see the use of SSL as
necessary, and even if major ISPs do support IMAP/SMTP over SSL (and I
think many don't) I suspect they would account for only a few certs
compared to the number of certs for web sites.
Before making that leap lets put things a little more into perspective
here, and of course you do realise how relatively few website
certificates there really are in the whole scheme of things right?
approx 100mil websites... or 20mill web servers and only 87,000 valid
certificates...
http://www.securityspace.com/s_survey/sdata/200501/certca.html
less then 0.1%... Unfortunately I can't find anywhere that does stats on
other protocols with/without encryption...
There are very few *comparitive* stats. The most
indicative is SSH v. secure telnet, and the latter
practically doesn't exist any more, as nobody wants
to futz with certs when SSH installs out of the box,
and raw telnet is only used for diagnostics of servers
these days. But, that's anecdotal evidence, in that I
haven't come across a secure telnet installation for
about 5 years, and I don't know any security people
who admit to using it.
In email, takeup of both systems - OpenPGP and
S/MIME - is down below the noise level. OpenPGP
is unsuccessful because the client isn't easy to
acquire, and S/MIME isn't successful because it
requires a cert. Both are unsuccessful as each
other, in market terms, more or less, as indicated
by a steady stream of startups in the space.
I've been hunting around for 2 years, and that's
about it! Very little comparitive data to back up
the theoretical predictions.
In particular, we might want to consider different security requirements
for the HTTP/SSL case vs. the IMAP/SMTP/LDAP over SSL case (e.g., I
suspect phishing is not that relevant for the latter case), but we're
forced to treat these two cases the same because they *are* the same as
far as both the CAs and the implementations are concerned.
Gotta split the security model up into more then just on or off...
I think there is a problem here - it isn't possible to
consider a security model without consideration to
the application. That is, if one was to use the
security model from HTTPS for the S/MIME system,
that would be wrong security design. Security
models cannot be divorced from the application.
So if you are considering security, the starting
point has to be to define the application. This
defines the threats. And that defines how one
should defend against the threats, in an economic
security model.
So getting back to rootlists, if a particular root list
is used for multiple applications, then that would be
a problem. Theoretically speaking, of course, and
if CAs were accustomed to finessing this issue, then
there might be logic in the software manufacturer
following suit. But, this does not obviate the basic
fact that the application is responsible for security,
so Firefox has to consider its model, and Mozilla has
to consider its model. They might be different...
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
