> I never suggested there was no incentive in better security. Treating
> unauthenticated encryption (which you call security), in the same manner
> as SSL with CA authentication, then all we have done is reduce the
> default overall level of security, from very high to nil IMO (or "self"
> as you call it).


Oh, no, I'm not saying it should be treated as
the same.  This is yet another reason why we
need the chrome - so as to give valuable
distinction between the various grades of
crypto based security.

The chrome branding box would simply state
in this case that it was an unauthenticated
self-signed cert connection, in some garish
colour, or monochrome, even.  It should be
entirely obvious to users - that is part of the
branding.


> I would strongly object to any kind of congratulations .


I present that more as counterpoint, to establish
how oddly placed the warning is.  In fact, the job
of the browser should be to not pass judgement
on the nature of the crypto connection at all - in
this sense, the Lock is a double edged sword, as
it does pass judgement (and hence we are led
into the whole "one size fits all" difficulty).

Rather, the job of the browser is to present the
info for the user to use, as they see fit.  Hence,
the branding box, stating the brand of the CA, or
in the case of a self-signed cert, conspiciously
failing to present...  Nelson suggested that it
display a demon for self-signed certs, although
the BSD demon is a watch word for security!

> If you wanted to eliminate warnings for self-signed certs, the only way
> I wouldn't object to it would be to treat those type of connections
> exactly the same as unsecured connections . However, that's not
> possible, because mostly they use the https protocol in URLs , which
> client applications today use to check for CA trust . You would need to
> come up with some new trigger for your unauthenticated encrypted
> connections (perhaps a new protocol handler) .


Well, really, what should perhaps be done is
for the HTTP (not https) protocol be capable of
bootstrapping to an ADR or a self-signed.  But,
the reason for not doing that is that it would
change too much of the protocol, and require
massive coord and agreement.

The changes that are being suggested are all
easy, small, and require no coordination.  They
also all positively benefit, especially in
concert.

iang

PS: ... That's why we've settled on them.  I've
been having this debate with dozens of people for
the last year, mostly over on the general crypto
groups.


_______________________________________________
mozilla-crypto mailing list
[EMAIL PROTECTED]
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to