On Mon, Mar 24, 2003 at 03:31:25AM -0600,
meor(_at_)mail(_dot_)SoftHome(_dot_)net wrote:
For those of you who have tossed around the idea of a hash-stamp or other
type of stamping protocol, I implore you to check out this proposal. I
believe it to be very close to a workable solution, if not one, for
limiting any type of undesired network communication, but specifically
eliminating SPAM E-Mail. I know I've thrown this idea out once before but
I only received a few responses from people who completely read the
proposal and understood it's workings. A web page has been put up
explaining the idea at http://meor.xwarzone.com/overview.htm For those of
you that have read it before, a few paragraphs have been added to the
beginning of the proposal to more properly introduce the concept and make
it easier to follow.
Doesn't seem to be anything new in this system or anything that solves the
major problems of such systems.
In fact, it seems to miss out on the goal of being able to pre-stamp a
message before you send it to a stranger, rather than needing a
challenge/response protocol on top of the test, which has a number of pitfalls.
On top of this, this system seems to be expecting some sort of PKI to be
in place for authentication. Or if there is no PKI, I presume it's of the
"First person to declare a key for address X sets the key" form. While such
schemes have their merits, they are not without some significant flaws,
including
the famous nasty denial of service attack (and now you have people motivated
agains the system, unlike mail encryptors, so DoS is a serious worry) and
the problem of people who lose their keys.
In general, while I was the first person in the community I know of to advocate
stamp systems, I now disavow them. A variety of the reasons are outlined
at www.templetons.com/brad/spume/estamps.html at the bottom.
I do agree that cpu-stamps are superior to money based stamps for a number
of reasons, but this system doesn't seem to do enough to take advantage
of those superiorities.
If you want to do a CPU stamp system and have any chance of its adoption,
what you need is the following:
When the system decides to reject mail from an unknown due to lack of
stamp, the "postage-due" notice should contain a URL to a java applet,
the URL containing full parameters to identify the message.
The message should be held in a spool.
The applet should calculate the response to the hard-problem. Since
the JVM is probably slow at this, a problem thought of as a 4 second
problem might well be a 40 second problem, however...
The applet should do its calculation in the background, in a window
that is iconized. It should pause frequently so that it does not
saturate the CPU of the user. Thus it might take even minutes to solve
the problem.
However, when the applet is done, it returns the calculated value to
the server from which it was downloaded, and then silently terminates.
The user does not have to be aware of this.
The server receives the calculation result and forwards it to the mail
system that issued the postage-due. The mail system checks the
result, and if it is OK, fetches the held mail from the spool and
delivers it. If not, it can be deleted or moved to another queue etc.
This is the standard of ease of use that a stamp system has to live up to.
Simply click on a link and forget about it, and your mail will be delivered.
However, even this has flaws:
a) It requires a JVM (or other fast portable executable platform. CLR could
also be a choice.)
b) It can't be done offline
And it has many of the other flaws of e-stamp systems. Mailing lists remain
a perpetual problem. For one, it would be a violation of the standard to
send any challenge to mailing list mail, ie. that which is tagged with a
Precendence: bulk header.
I believe the best solution must permit a global whitelisting of legitimate
mailing list operators, though it can be done through a number of means.
In a way, Habeus is a very basic form of that.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg