Non-mail-specific standard (proposal)

Hallo,

I think that it would be a good idea to have two standards:

1. A general-purpose specification that deals with the encoding (in a
way compatible to domain names), with the following encoding procedure:

  . for each label passed to the encoding function, do the following:

    . split them at certain characters (delimiters) into sub-labels
      (i.e. punctuation without "-").

    . for each sublabel, do the toASCII function as defined in IDNA.
      Possibly with mixed-case annotation

    . for each delimiter, do a Unicode normalisation.
      Also map the ideographic full stop (wide and narrow) to ".".

    . re-combine the sub-labels using the normalised delimiters.

2. A standard that says how this is implemented in mail:

  . everyting allowed by RFC 2822 but not by RFC 2821 is considered
    part of the embedding, not the address.

  . labels are:

    . the local-part (with minimal quoting, not dequoted)
    . the individual domain labels of the RHS or the string between the
      '[' and ']' (as an optimisation, one can also just use the RHS).

  . the '@' has to recognised in all its variants.

  . if the domain output of the encoding function of the local-part is
    identical to the input, the local-part is copied directly from the
    input (so that there's no quoting minimisation introduced).

Other standards and config file parsers could use the first
specification, too:

  . Netnews:

    . labels are the segments of the newsgroup names (again, as an
      optimisation, one can also use the ng name as a whole).

  . Bind:

    Yes, it would produce the right result for "J\\ohn\.Dö".exämple.com
    (the quoting might be correct here as I don't care about bind's
    quoting rule, important is what follows):

    labels as recognised by bind: <<"J\\ohn.Dö">>, <<exämple>>, <<com>>.

      for first label:
        sub-labels/delim's: <<">>, <<J\\ohn>>, <<.>> <<Dö>>, <<">>
        encoding of them:   <<">>, <<J\\ohn>>, <<.>> <<xn-D-xxx>>, <<">>

    encoding: <<"J\\ohn.xn--D-xxx">>, <<xn--exmple-xxx>>, <<com>>.

    email address: <<"J\\ohn.xn--D-xxx"@xn--exmple-xxx.com>> (same as
    result procuded by IMAA).

  Domain Names:

    For valid domain names, all of these produce the same result:
    . IDNA, labels joined by '.'.
    . Spec #1 with the domain name passed as a whole.
    . Spec #1 with the domain name as individual labels, joined by '.'.

Claus
-- 
http://www.faerber.muc.de/