[Mimedefang] Reputation Reporting Protocol submitted to IETF as an I-D
kd6lvw at yahoo.com
Sat Jun 19 03:02:34 EDT 2010
--- On Fri, 6/18/10, David F. Skoll <dfs at roaringpenguin.com> wrote:
> I've submitted the Reputation Reporting Protocol ...
I can say this right off: Encoding the timestamp as an integer with an anchor year is going to be a problem (e.g in 2038 or 2106) requiring a future version upgrade. Consider a split timestamp: One value in seconds since the beginning of the current year (25 bits required), and another for the cardinal number of the year itself. A 32-bit year as a separate value should last about as long as the Earth itself is expected to.
For the extra 7 bits in the seconds value (as a 32 bit value total), one could encode 3 of them for the day of the week as a "verifier" value.
Maybe you don't care about the 27-year timebomb you're giving yourself.
IP-address-types: Consider adding as a separate value "spamtrap" that occurs as a result of delivery to a spamtrap address. Such may be classified without passing through a filtering system. It is a case that should be distinguished from #3 (filtered auto-spam) or #7/#8 ([in-]valid recipient).
Section 8 - Example Report: Should use an IPv6 address from the documentation prefix (2001:DB8::/32) instead of a live address.
Those were the only three things I saw worth commenting about.
More information about the MIMEDefang