Re: [Fwd: Request for Advice on VGRS IDN Announcement]

[Thomas Narten <narten@us.ibm.com>]

> 	If one asserts that things are so underspecified that the MS
> behaviour is not invalid, then the only logical conclusion is that
> the same "under-specification" means the Verisign behaviour is not
> necessarily invalid.

Two wrongs might make  a right? :-)

Actually, I think there are two separable issues: what one sends out
in a query, and how one inteprets the query.

It's a bad idea to put UTF-8 in a query, but it also doesn't seem to
be a clear violation of the spec.

But the spec seems to be pretty clear (from my understanding) that on
comparisons when doing lookups at the server, there is no room for
proprietary intepretations. One should just do a byte-by-byte
comparison (after doing case folding if possible).  Anything else
(even if the querier was sending out stuff that it probably shouldn't)
would only multiply interoperability problems. And the idea that one
might in essence fabricate responses for all queries that have the
high-order bit set raises interesting questions about required server
changes to make this happen and whether all DNS servers that are
authoriatitive for a domain have to implement the same algorithm. I
would think so (all servers should respond identically). But doing
this in the absence of an agreed to standard raises all sorts of
intersting questions/implications, a number of them political.

Thomas