M-label definition

[John C Klensin]

--On Tuesday, April 07, 2009 16:59 -0700 Lisa Dusseault
<lisa.dusseault at gmail.com> wrote:

> I'm not certain a new term needs to be introduced.  If we're
> talking about a string that is invalid as a label, giving it
> that term seems to legitimize it.  If we're talking about a
> string that may or may not be valid, that's just "a string"

Sorry for not including enough context in the note.

One of the big advantages of the U-label / A-label terminology
is that a U-label is precisely what one gets when one converts
an A-label back.  The term is obviously not defined for
IDNA2003, but the IDNA2003 definition would require that a
U-label be the result of applying ToASCII to something valid and
starting in "xn--".

That definition of a U-label as being the thing that can be
translated to and an A-label without any loss of information, is
critical to a number of pieces of the draft specs.  It has also
become fairly widely used in the outside community, almost
certainly to the point that we would be doing the community a
considerable disservice by trying to redefine it (that is why
Yao's suggestion does not work).

When we first started talking seriously about including some
mapping as part of IDNA2008, Patrik suggested that we define a
separate term, M-label, to distinguish between U-labels (which
could be recovered from A-labels) and strings that could be
mapped into them (which could not).  Mark made a very similar
suggestion a short time later.  But their definitions, at least
as I understood them, were different.  Patrik appeared to be
suggesting that they be disjoint, s.t. the set of U-labels and
the set of M-labels together would make up the set of non-ASCII
strings that were valid IDN labels.   Mark appeared to be
suggesting a subset relationship such that the set of M-labels
were, identically, the set of non-ASCII strings that were valid
IDN labels and U-labels were the subset that could be translated
back from A-labels.

Hence the question.

    john