From: Bill Stewart <stewarts@ix.netcom.com>
Date: Mon, 31 Mar 1997 22:57:29 -0800 (PST)
To: cypherpunks@toad.com
Subject: Re: remailer spam throttle
In-Reply-To: <199703281917.TAA03495@server.test.net>
Message-ID: <3.0.1.32.19970331225622.00633a98@popd.ix.netcom.com>
MIME-Version: 1.0
Content-Type: text/plain


At 06:46 PM 3/28/97 EST, Dimitri wrote:
>That's a good idea, but it'll take up a lot of disk space at the
>machine running the remailer.  Right now, remailers that provide
>latency don't keep an e-mail for more than about 12 hours. Once
>you start keeping them around for a few days (a reasonable grace
>period for a first-time user), it's a lot more disk space.

Typical remailers carry maybe 100-500 messages/day;
typical messages run 1-20KB unless they're pictures or warez.
	(Yes, I'm making these numbers up....)
That's about 1-10MB/day of traffic.  If you keep it 7 days,
that's up to 70 MB storage, which you'd probably keep in /tmp
to avoid backups and maybe avoid disk quotas.
If you're running it on your own desktop machine, that's small;
it's a bit large for a laptop, and whether it's reasonable for an
ISP shell account depends a lot on your ISP's policies and disk quotas
(and /tmp clearing policies.)

>Suppose a LEA wants to search the computer hosting the remailer.
>They come across a bunch of encrypted files.
>The operator has to convince the LEA that they don't have the means
>to decrypt the e-mails or even to figure out who they're from.
>That just may be close to contempt of court. 

But you don't have to explain it to the LEA - you may have to
explain it to a court, but you get to bring along a lawyer.
It's not contempt if you're telling the truth (that you can't decrypt it.)

On the other hand, if you're using a two-part key (one sent to the
recipient, one kept in transient memory on your machine plus in your head),
you _do_ get to explain to the judge why you're refusing to testify,
and why the ECPA protects the privacy of all the messages that
may be revealed if you release your half of the key,
and why you need sworn testimony from the LEAs about exactly which
messages they've eavesdropped besides the ones addressed to their target,
and about why giving them the key would violate the privacy of
the recipients of those other messages so they can't have it,
but you also refuse to decrypt the ones for the victim yourself,
at least without a direct court order (as opposed to a mere warrant) -
_then_ you'll have the opportunity to get close to contempt of court :-)

>Say, you might be asked to 
>explain how you generate the "random" keys so they can be recreated.

If the "random" key includes a part you know as well as a part
the user knows, and the user's part not only includes a hash of
the message (so you need to know the contents of the message to
recreate the session key) but also the usual things like the
system clock to the microsecond and the contents of the rand pool,
and maybe a few hits from /dev/random - then it's perfectly fine
to tell them.

>IMO, the 'net has changed from what it used to be a few years ago.
>One can no longer send e-mail to an unknown recipient and hope that
>they're willing to accept anonymous e-mail. 
I'd agree, but from the first anonymous remailers open to the public
there were people who didn't like receiving anonymous mail :-)

>unless the remailer knows that the recipient took some positive
>action to indicate that s/he has a clue (such as, added a key to a
>keyserver), their anon mail should be immediately discarded and
>they should instead get a note:

That's an interesting approach - a bit extreme, but the main cypherpunks
applications for anonymous remailers are things like whistleblowing
(which can be posted to the net or emailed to people like Foo Inspectors
who _ought_ to be willing to accept anonymous mail) and potential
co-conspirators (who _ought_ to be willing to accept it if they're
interested in co-conspiring), and of course yourself under various aliases.

>> > Right now, there's a very large number of addresses in the key servers.
>> > Instantly making them into a list of addresses that accept anon mail
>> > will make it hard (hopefully infeasible) for the LEAs to investigate
>> > everyone willing to accept anon e-mail as a suspect in sending it.

A nice touch.

>> To solve this problem you need to do a ping message, "please reply
>> with this nonce to be blocked".
Yep.


#			Thanks;  Bill
# Bill Stewart, +1-415-442-2215 stewarts@ix.netcom.com
# You can get PGP outside the US at ftp.ox.ac.uk/pub/crypto/pgp
#     (If this is a mailing list, please Cc: me on replies.  Thanks.)