--On Thursday, May 09, 2013 09:28 +0200 Randy Bush <randy@xxxxxxx> wrote: >> Similarly, "wherever possible" is unhelpful; if it's not >> possible to fully-qualify a domain name then ambiguity is >> guaranteed. > > no, that is what SHOLD means. e.g. when i write docco that > has an ops clause where there is likely a knob or other op > choice, i do not think i have the prerogative to tell the op > how they MUST run their network. but i can say that they > SHOULD do x. If you are writing a protocol spec and you believe various bad things can happen to you if partially-qualified names are used, sure you do. Whether your belief about possible damage and its likelihood is reasonable is another question and that, IMO, is the discussion we should be having. Now that op has prerogatives too. Because conformance to IETF Standards is voluntary, it can ignore the MUST, leaving you the option of calling the protocol police and/or deciding not to deal with that op. Or it could decide to not deal with you because you make such unreasonable demands. But neither of those options deprive you, or the IETF, from saying "MUST" if you are convinced it is necessary. IMO, "necessary" should include interoperability, security, or stable and predictable operational reasons, despite 2119's apparent restriction to the first of those. best, john