"Eric A. Hall" <[EMAIL PROTECTED]> wrote: > You can pass a stringprep profile to ToASCII as a parameter, but an > empty parameter value (or some other flag) means that the string has > already been prepared and that steps(N) are to be skipped. Would that > work?
That would be a fine way to implement it. In the specification, it needs to be clear that "apply ToASCII" means the proper Stringprep profile must be applied, it is not optional. Exactly when and by whom it is applied is an implementation detail. > Clarification, the resolver doesn't "need" a flag because it already > has access to a flag. On the application side, this would be the Wide > API versus the legacy API That works until you have a name in which some labels are non-ASCII text and some labels are opaque bytes. I guess in that case the application could just use the legacy API and do the IDNA conversions itself. But another solution is for the new API to take an extra bit per label as input. AMC
