Oki all, I was interested in a policy I came across recently at a cctld registry. If a domain has no (or few for some value of few) hits over some period of time post-registration, the registry will recover the string and let another user acquire it, and presumably actually use it. So if t = 3m, pokey.cctld could go to four users in the course of a single year, iff the first three made insufficient use of pokey.cctld during that time. I'm going to guess that panix.com is different from most of the multi-k domains that are dropping off the VGRS registry and into today's (well, yesterday's) drop pool, as measured by use. I'm going to guess that panix.com is different from most of the multi-k inter-registrar transfers of yesterday, today, and tomorrow, by the same use metric. IMHO, organizing policy around function, actually distinguishing between panix.com and the overwhelming majority of domain names for which some change of state at the registry occurs, is a better principle that to continue to organize policy around trademarks and their buyers and agents, indifferent to the frequency and distribution of use of a domain name. At some point, it really _is_ a name-to-addresss map, and not a cognate for a trademark-to-owner map. It is possible to distinguish risk, and a policy which chooses not to make distinctions isn't prudent. In case anyone's missed the obvious, we now have an incoherent dns, and caching resolver operators have introduced the incoherency, and no one in the operator community is visibly spitting blood at the intentional exception to rfc2826. This situation should not continue. Neither should the "new/hijacked" answers be served. Eric