On Sun, 11 Oct 1998, Liviu Daia wrote:
On 10 October 1998, Aaron Schrab <aaron+pf(_at_)schrab(_dot_)com> wrote:
Of course messages should go to the address that you used to subscribe
to the list. But, for a lot of people it would be good if either form
would work for sending messages to a restricted list, or possibly for
unsubscribing.
You miss the point. Assume that, for some weird reasons, I want to
subscribe with both addresses. Assume also that, a few weeks later, I
want to unsubscribe "daia(_at_)imar(_dot_)ro", but keep receiving messages at
the
first address. With the patch, both operations won't do what I expect.
That's not true. Even with the patch, multigram is able to tell the
difference between the two addresses:
2 daia(_at_)imar(_dot_)ro 32749 daia(_at_)imar(_dot_)ro
1 daia(_at_)stoilow(_dot_)imar(_dot_)ro 26195
daia(_at_)imar(_dot_)ro
1 daia(_at_)stoilow(_dot_)imar(_dot_)ro 32728
daia(_at_)stoilow(_dot_)imar(_dot_)ro
2 daia(_at_)imar(_dot_)ro 26195
daia(_at_)stoilow(_dot_)imar(_dot_)ro
so you wouldn't need to be concerned about unsubscribing the wrong one.
The unpatched multigram still reports them as similar, it just gives a
much lower score:
1 daia(_at_)stoilow(_dot_)imar(_dot_)ro 18813
daia(_at_)imar(_dot_)ro
Automatically aliasing addresses that are not the same is IMHO a Bad
Thing [TM], no matter what noble purposes you have in mind when you do
it.
The multigram algorithm isn't the same as automatically aliasing them.
There's no question that cryptographic cookies are preferable when real
authentication is desired, and that unsubscribe is often one of those
cases; but in many circumstances they're overkill.