Re: [idn] WG last call summary

[John C Klensin <klensin@jck.com>]

--On Tuesday, 19 March, 2002 12:29 -0500 "J. William Semich"
<bill@mail.nic.nu> wrote:

> Correct for RFC 1035 and the BIND 9 implementation of it at
> least.

Well, we can keep repeating this until it becomes true, water
flies uphill, and other convenient things happen.  However, a
careful reading of RFC 1035 is illuminating:

(i) It requires that all character strings (explicitly including
labels and domain names, not just "host names") be compared in a
case-insensitive manner (section 2.3.3).  As this working group
has discovered multiple times now, case-insensitivity is not
precisely defined for many non-ASCII scripts.  And, in
particular, there is no simple operation (i.e., not requiring
mapping tables) that implements case-insensitive mapping for
arbitrary UTF-8 strings.

(ii) While it provides that "non-alphabetic codes must match
exactly" (end of section 3.1), it would be quite a stretch to
argue that the characters of, say, Greek are "non-alphabetic".

In any event, one can't just put binary information in there
without knowing whether it is to be compared case-insensitively
(whatever that means) or absolutely.   And RFC 1034 strongly
implies that binary information can be used only with new RR
types (or, presumably, with new Classes, not in the existing RR
types and classes.

      john