on 6/10/2002 8:12 PM Dave Crocker said the following:
> The IDNA effort is about domain NAMES. Yes > As nearly as I can tell, this thread is about domain name PARAMETERS, such > as non-name fields in RRs. No RRs define the syntax of their owner name as well as the syntax of the RR data. New RRs often need new domain name syntaxes, either for the owner name, or for the RR data (although it's hard to imagine an RR data domain name which isn't subsequently used for an owner name at some point). The rules currently defined in IDNA impose restrictions on the syntax of every new RR invented from here on out. While these rules are appropriate for the owner names of legacy RRs which are queried through direct input, they impose unnecessary restrictions on every subsequent RR which could ever be defined, and regardless of where the input comes from. This is a fundamental architectural design decision which needs very serious evaluation. It seems to me that it is an easy problem to avoid: decouple nameprep from the codec and the problem goes away forever, instead of hanging around forever. Nobody has told me why this is undoable. Feel free to weigh in. -- Eric A. Hall http://www.ehsco.com/ Internet Core Protocols http://www.oreilly.com/catalog/coreprot/
