From: Russ Allbery (rra@stanford.edu)
Date: Wed Apr 02 2003 - 18:04:33 CST
Matthias Andree <ma@dt.e-technik.uni-dortmund.de> writes:
> Russ Allbery <rra@stanford.edu> writes:
>> You folks *really* need to go read the news.software.nntp discussion
>> about xBin. The reason not to use Path, or for that matter anything
>> other than message ID, is because you want something that you can
>> filter during CHECK/IHAVE, and relying on your upstreams to pre-filter
>> before sending you the feed is a nice idea in theory and a lost cause
>> in practice. Like it or hate it, and I mostly hate it, using the
>> message ID solves practical real-world problems in efficient and
>> significant ways, and the other proposed solutions kind of... don't.
> So is your suggestion to register and reserve a special, invalid
> top-level domain name such as "binary" so it can be included in the
> Message-ID?
I'm not suggesting anything at all. I'm pointing people at a much more
comprehensive discussion of this. Curt Welch, who was the person who
actually proposed xBin, has already reserved a domain for that purpose and
written up most of a protocol, and has posted quite a few details on its
purpose and use. The basic idea makes me cringe, but I have no good
counter-arguments against it, and there are ways in which it is obviously
useful if used and would solve real problems, so I don't really have an
opinion on it one way or the other.
> It doesn't solve Jürgen's "may be forged" issue: spammers will just omit
> the ".binary" tag.
xBin addresses the "may be forged" issue in rather impressive detail,
using a hierarchical hashing system.
> And it will take quite some time to become efficient -- the problem with
> the Message-ID is that you can't change it later (so you can't just add
> a .binary tag when you figure the message is binary -- you'd cause dupes
> that way).
Yes, the poster has to use xBin; you can't retrofit it after the fact.
-- Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>