On Mon, Apr 16, 2001 at 02:42:51PM -0700, Russ Allbery wrote:
Brad Templeton <brad(_at_)templetons(_dot_)com> writes:
Yes, but manual verification of signatures is of dubious value.
That seems to be a non-sequitur. It's certainly possible to automatically
verify a multipart/signed message. Some news readers and many mail
readers already do this.
If all the software does is include a string about the verification, that
is then manually checked, it's manual verification. Humans ignore
warnings which happen frequently. Thus if most (or even just a lot, like
a few percent) of otherwise valid messages have the warning, people learn
to ignore the warning.
For digital signature authentication to work properly, it is
unfortunately necessary that all messages in a class be signed, and that
the presence of an unsigned or improperly signed message be a major
anomaly which gets a fair bit of attention, or which is in fact
forbidden.
You're solving a different problem than I'm talking about. For occasional
use for important announcements, the multipart/signed protocol works just
fine. I understand that you're trying to solve the problem of fully
authenticated message streams, which I agree is an interesting theoretical
What are these "important announcements" though? As I indicated, I mean
that all the messages in a "class" have to be signed, so that it becomes
worthy of note when a message in a class is not signed, and this can cause
a warning that will be paid attention to.
So what is this class called "important annoucements" and how is it defined,
and how does the software detect it, and know when an important annoucement
comes that is not validly signed, so that it can warn me about it?
If you just mean that "before you say something important" be sure to sign
it, again the value of that is limited, though not zero. It means that
people who are in the know, after reading something of some importance
(whatever that is) would then look to see if it is signed, or have their
brain pay attention to the "this is not signed" warning which appears on
almost all messages.
I would rather solve the real problems the net faces with security,
which are unauthorized control messages (and the resulting lack of trust
in authorized control messages), most of all cancel, and large scale spam.
Don't get me wrong, I think bodies should be authenticated, but the value
from it, at least in today's net, is minimal, and the backlash over it,
namely a lot of mime stuff most readers don't know how to handle tacked on
the end of messages, far worse than the benefit.
We will eventually need to put all that mime stuff in, and it will annoy
a lot of users (we've all seen this) but we should do it for something
really useful, not mostly ignored signature of bodies.