The problem that got us here was that registrars have historically been not flexible enough at releasing domains when the owners *did* want to transfer them.

George, The point I tried to make in my prior note was that not all domains have the same temporal property of non-functional change. The "problem" that you refer to exists for some domain owners. Bruce asked for the comments of this subscribers to this list, on the current ICANN transfer process. Since ISP/NSP/... change registrars (cosmetic non-functional change for a cost savings of $0.10/day, maximum) almost never, it is wicked unlikely that the authors of the current ICANN transfer process ever thought about network infrastructure operators as affected or interested parties to any policy change. "We" didn't have "the problem", historical or otherwise. With the exception of operators who's business value is organized around resolution in under 3 days for new customers, not ongoing resolution after the 3rd day, or who's business value is now organized or re-organized around resolution in under 2 hours with the new dynmaic update property of several registries, and not ongoing resolution, "we" have been pretty much problem free in the registrar and registry space since Jake Feinler and Jose Garcia-Luna ran the SRI NIC. If webhosting outfits want to bundle registrar-reseller into their package forcing registrar transition with renumbering, fine. But they are further down the food chain. If the registrars want to directly slam the end-users, that's fine too. But short-term 1U renters and vhost operators and registrants aren't the NANOG list, and that's what Bruce asked, cosmetically or otherwise, for input from. An unintended side-effect of "competition" between registrars is that the named network infrastructure is someone's target of opportunity. In his reply to my note, Bruce points out that the system works for all. There are two classes of domain names already. Registry reserved and not. Adding a record to the database, or a lookup in addition to the existing access, to implement a third class, could get the domain names associated with critical network infrastructure out of the risk pool for whatever the transfer model de jour is for registrar competition, and make "rollback" for this class technically distinguishable, therefor policy differentiated, from the general zoo. Why don't you collect the results of a survey of access ISPs and above who change their own domain names registrars more than once every five years and show me that NANOG is equivalent to vhost-interest@apache.org. Cheers, Eric