mappings-01

Paul Hoffman phoffman at imc.org
Sat Jul 4 17:10:32 CEST 2009


At 8:40 AM -0400 7/4/09, Vint Cerf wrote:
>second, do the two of you have a preference for how this mapping 
>should be referenced?

Yes: if it referenced with RFC 2119 language, that language should actually match the definitions in RFC 2119.

>Many participants have suggested that the mapping SHOULD be applied so 
>as to convey
>the importance of the regularity the document produces but allowing 
>for circumstances that
>might justify not mapping. I would underscore the weight given in IETF 
>terms to "SHOULD"
>as one is expected to adhere to the advice and only deviate for very 
>good reasons.

As Pete said, we *took out* the RFC 2119 language from this draft. When a document says "you SHOULD do xyz", it ought to say when the SHOULD needs not be done. Another way of saying this is, for every SHOULD, the document ought to say why it was not a MUST.

Everyone unclear on this should go back and read RFC 2119. It is quite short, and quite definitive.

In the case of mapping user input, we could not give a good reason why this is needs to be required. It is clearly a good idea because it will prevent user surprise, and we say  so. However, mapping does not promote interoperability between DNS clients and servers, nor between applications. Things that are just good ideas where the exceptions cannot be well defined are not, in my opinion, applicable targets of RFC 2119 "SHOULD".

If the WG has consensus a proposal for the exact wording they want that uses RFC 2119 language that actually meets the requirements of RFC 2119, Pete and I would be happy to incorporate it. We suspect that it won't happen, and that we'll waste a long time arguing about it, which is why we ripped out the RFC 2119 language in this round.


More information about the Idna-update mailing list