"Peter J. Holzer" <hjp-asrg(_at_)hjp(_dot_)at> wrote:
On 2005-01-07 20:02:31 -0500, Seth Breidbart wrote:
for a sufficiently large ISP (think AOL, or a cable modem business),
keeping track of the tuples of <sender, recipient> for any length of ti=
will require a _substantial_ database infrastructure.
i was going to do some back-of-envelope computations on sizing, but
this is your proposal, mitch, so why don't you do them and share the
results with us.
Under 1K of data. Remembering stuff is doing it the hard way. Create
the lhs of the Message-ID signed with a private key that changes
That assumes that the Message-ID is created by the MTA. Many MUAs create
the Message-ID themselves, and the MTA should not (indeed must not, per
RFC 2476 and 2821, if it is syntactically correct) replace it, because
that would make the "In-Reply-To" and "References" headers useless.
You're right. They'd have to remember it in that case. However, AOL
(specifically) wouldn't have that problem.
They can also play games either with md5 or hash tables, so the amount
they have to remember is a lot less. It doesn't matter much if there
are a few false positives.
Asrg mailing list