Hi Heikki,

> I agree, and in fact there is some relevant prior art/code to get
> started. Gaim has an encryption plugin that uses NSS. The plugin can
> generate a certificate for you when you start. Then when you get sent an
> encrypted message, it checks if you have conversed with that peer cert
> before, and if not, will ask the user. From there on, as long as the
> cert your peer presents to you remains the same things just happen
> automatically.


How does the 1st encrypted message get sent?  How does the
cert get sent?  User action?  Hints in received messages?


> A thing to be cautious about with automatically generating a certificate
> and perhaps automatically signing all messages (and encrypting when
> possible) is that sometimes you want to remain anonymous (sending mail
> through anonymizers etc.) In those cases sending even one supposedly
> anonymous message as signed message can be a disaster. Might be better
> to err on not signing by default, and sign+encrypt by default when
> corresponding with a peer whose cert you have.


It all depends on what you mean by signing.  If
signing reveals any more than "this message came
from this cert" then yes, that's an issue.  If it
is simply a message authentication code, then how
do you intend to authenticate the messages without
signing?

My view would be to drop signing as being an
available operation of human meaning.  I.e., do
it transparently if at all, for purposes of tying
the message to the source cert identity, only.  If
a "human meaning signature" feature is required,
add that over the top of the protocol, at the
application level.

Having said that, simply dropping the authentication
entirely and encrypting without any auth technique
would be an entirely valuable thing.  It's still
wonderfully more secure than plain text email, and
experience from the OpenPGP email community (which
could be considered substantial) is that any active
attacks on email users would seem to be remote and
optional, even in the face of known attacks.

I.e., leave all sigs for another day if one strikes
difficulties.  Message source and integrity checks
should be a nice to have.

As to your comment on anonymous email - I'd suggest
this is a specialist requirement.  Don't include
anything in there that might lead us down a blind
alley.  Most people most of the time do not want or
need to send anonymous email.  Let's not break the
app for the majority just to please the anarchists.

(speaking as one of the latter, myself :-) .

iang
_______________________________________________
mozilla-crypto mailing list
[email protected]
http://mail.mozilla.org/listinfo/mozilla-crypto

Reply via email to