Eszett

[Andrew Sullivan]

On Wed, Aug 13, 2014 at 06:07:18PM +0000, Shawn Steele wrote:
> to remove the mapping (so it’d get it’s own punycode).  However most
> browsers use the UTS#46 transitional rules because that would allow
> a link that previously went to one server to end up going to a new
> server if IDNA2008 rules were applied.

I disagree.  In the case of Eszett, what UTS#46 gives is "IDNA2003
behaviour", not "transitional rules".  There's nothing in UTS#46 about
how to sunset this behaviour, the timeline for the changeover, how it
would work, or anything of the sort.  What it does is increase the
length of time during which the mapping is preserved.  We still need a
flag day to cut over in the future.  I would be delighted if UTS#46
provided a workable transition strategy.  But I can't see how it
actually achieves such a goal as currently written.

> IMO I’d really like to see a “display” record so that if someone
> wanted their display name (that equivalent to the IDNA canonical
> form) look a certain way, then they could for presentation.

Except, of course, that no actual browser would put up with the
latency induced by an extra second round trip to get that record, so
there's no hope of it being used.  If it's not going to be used, then
it won't be added to zones either.  So it won't be deployed, therefore
nobody will specify it.  Worse, it won't work consistently, because
sometimes when you want to format something for display (when writing
in a word processor, for instance) you actually care about this and
_would_ put up with the latency, except that you might not be online.
So the display of the string will bounce around depending on whether
you are online or not.  That doesn't sound very satisfying.

Remember, too, that on the first day of deployment of that record,
virtually no clients will have it, so zone operators won't have an
incentive to put the "display" record into the zone.  Therefore, any
browser that wants to use the record will do a lookup that is all but
guaranteed to return nothing.  In my experience, browser developers
simply won't go for the extra latency for no benefit at all.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com