RE: CAPTCHA over smtp (yet another spam solution to discuss)

Joachim wrote:

> Seth Goodman wrote:

<...>

> > 1) Read-only IMAP is a more modern version of news.
>
> News (NNTP?) is /not/ read-only.

You can post but you can't erase a message, as you can in your private
message store.  But you are correct, it is not read-only.


> > That's been around a very long time,
>
> "That" == News?

Yes.


> > but I'd still guess that more people read them via email than news.
>
> "them" == mailing lists?

Yes.


> > For better or worse, bandwidth is less of an issue than convenience
> > for many people.
>
> I think you may be saying, "it's more convenient for users to
> subscribe to mailing lists by entering their email address into some
> web form rather than figuring out how to make their MUA subscribe to
> RSS feeds or to read-only IMAP folders."

Yes.  I can't make them want to do more than that.  You can't make me
make them.


> > 2) Posting a bond to an mailing list you wish to post mail to would
> > be seen by many people as exclusionary.  Several mailing lists that
> > I read are unwilling to even require posters to register, as that
> > is seen as too restrictive.
>
> My point again:  Make all restrictions configurable, for the
> recipient. Where in this case, the recipient is the mailing list.

Then the folks who today want no restrictions will prevent such a system
from being used.  They block that currently, and they don't appear to be
losing any ground.


> > Though I don't agree with them, I mention it to show there is a wide
> > variety of opinion on what is too restrictive and I doubt that
> > requiring a cash bond before posting would be widely accepted.
>
> It'd be okay if it were accepted where it matters.  Like for mailing
> lists with thousands of subscribers, which you bring up next...

These are exactly the lists who will be the last to do it.  I'm thinking
of Debian-user, which has somewhere around 5-10K subscribers.  The folks
there consider registering on a web form to be overly restrictive:  you
have to agree to get mail (due to limitations in their list software)!


> > 3) Captcha's as a barrier to mailing list posting would be even
> > less of a barrier than for ordinary mail.  You pay someone in a
> > poor country a few cents to defeat the captcha for a mailing list
> > that gets distributed to a few thousand recipients.  That's
> > irresistible.
>
> Administrators of large-scale mailing lists have an incentive to keep
> their lists non-polluted.  Let them decide themselves what to do with
> that incentive.  They might moderate, they might require bonds, they
> might want to have captchas solved.  Why impose a one-size-fits-all
> pseudo-solution?

I'm not proposing this solution.  They can charge money for posting
today with no special infrastructure.  Why are there not many lists
operating this way?


> Seth Goodman wrote:
>
> > > No; "repudiable" is correct.  The verifier sends a challenge, and
> > > computing a valid reponse requires the private key.  But the
> > > prover can repudiate ever having "signed" (in the offline sense)
> > > because it is trivial to manufacture, given any response, a
> > > corresponding challenge.
> >
> > I agree with Brian that there is probably some confusion here.  One
> > of the strengths of public key signatures such as GPG is that they
> > are non-repudiable.
>
> For an offline system, that's not a strength, it's a core feature.

That's right.  It was a requirement for a digital signature system.


> In an online system, the sender issues a zero-knowledge proof of their
> knowledge of a valid signature.  The recipient does not have the
> ability to prove anything to anyone, although they have the entire
> transcript of the communication.

If the assertion requires zero knowledge by the sender, then anyone can
make it.  Perhaps the only thing you can assert without requiring
knowledge is that the party at the other end of the wire is alive.  I'd
like to know a bit more than that before accepting a message.  However,
I don't belief it worth the bad feelings it generates to force senders
to solve on-line puzzles simply to distinguish them from machines.  I
have already refused to deal with C/R systems when it annoys me enough,
so I don't think it wise to inflict this on others.


> > GPG signatures do give all future recipients that ability, though I
> > would suggest it is unnecessary for ordinary email.
>
> It's not just unnecessary, it's a major impediment for sender
> authentication to work.  Even if PKI worked (think: identity-based
> encryption), even if MUAs came with built-in crypto support and made
> its usage easy, people will feel uneasy about non-repudiably signing
> their every words, although most won't admit it.

That's news to me.  If you don't want someone to trot out something you
wrote down, then don't write it down.  When I send someone mail of any
kind, I have to assume that it could surface at some later time.  It
matters little whether they can prove mathematically that I am the
author.  All that matters is whether others believe I wrote it.  If I am
not comfortable with the possibility that I will be faced with a written
record of my own words, I use the telephone.

IM is probably the worst in this regard:  you believe it is an
ephemeral, informal communication, but every user on the channel has
access to a timestamped log with every word.  Even worse, you are
deprived of the benefit of nuanced expression, as with voice, and the
time to review your words before sending, as with email.  I think it is
possibly the worst of both worlds for business use, though it is
wonderful among friends.  Just one person's opinion.

--
Seth Goodman