Re: [idn] Unicode tagging

[Keith Moore <moore@cs.utk.edu>]

> The TXT is no real problem. It need not be matched, but it needs to be
> normlised using form C so that everybody knows how to handle it.
> There are also text in HINFO that need to use the same encoding.
> In the SOA RR there are domain names that define an e-mail address, not
> a host name (they are not used in matching, but are returned as data).
> So not all labels can be normalised as form KC as data may be destroyed.
> And text only may neither use form KC as it destroys data.
> So there will be both form C and form KC in use, if we require host
> names to be normalised using form KC.
> 
> But we also have the SRV RR. Here the owner name, that is the name used
> in matching, is not a host name. And there may be other RRs. And the rules
> for them may differ from host names. But it might not be any problem for
> matching of names. We should only have one way to do matching of
> owner names, independantly of type of name (like host name, service name).

I think in this group we should discuss labels only, not the contents
of RRs.  Very few of the RRs have content which is intended for human
consumption.

SRV RR labels can use the same matching rules as for other RRs.
There is no "owner name" matching an SRV RR label, it's just 
the name of a protocol.  There's no pressing need to make these
internationalized, they're not visible to ordinary users.

I don't know of any particular reason why TXT needs to be normalized,
as long as people agree what encoding is in use.  The format of
email addresses is not within the scope of this group either, so
we shouldn't change SOA in anticipation of email addresses being changed.  

Keith