What registries might do (was: Consensus Call on Latin Sharp S and Greek Final Sigma)

Andrew Sullivan ajs at shinkuro.com
Mon Nov 30 21:07:15 CET 2009


On Mon, Nov 30, 2009 at 09:48:38AM -0800, Lisa Dusseault wrote:

> Instead of restriction, what if registries could be relied upon to
> bundle a PVALID ß with 'ss' for a period?  

Right.  That's clearly open in this case (it's harder -- painful, even
-- but not impossible for ς too).  In my own view, if these
previously-mapped characters become PVALID, then the best thing for a
registry to do would be to try to provide maps.  Of course, that may
be difficult.  For example, when I worked on .info, we simply didn't
collect the Unicode "intention" for a label that included only ß,
because that character was simply mapped.  (I had a database developer
from Germany who yelled at me practically every time he saw me about
the mapping, BTW, as though it was somehow my fault.  So I can report
that there are some -- "at least one" -- people who want ß to be
available.)  So registries might not be reliable about mapping, not
because they don't want to provide the service, but because they can't
for practical reasons.

> There's still a nuance where possibly the registry has not yet
> registered those ß as bundled with the equivalent ss domains, and I
> can imagine two ways of doing this.  One would be to suggest a start
> date for registries to respond to requests for ß domains with all the
> appropriate ß/ss bundled domains, and that's the date when browsers
> can start to upgrade behavior.  Another way would be for browsers to
> look up ß first and fall back to mapping.

Well, remember that "registries" here means "every zone operator in
the world".  I have pretty grave doubts that we're going to be able to
co-ordinate what they all do.  Remember, too, that for many operators
the policy is "what I typed into the zonefile" plus whatever happens
when they do `apt-get upgrade bind` or something of that nature.

It does seem pretty wise, however, to suggest that some mapping of
this sort be planned with respect to the probable user community.  In
John's defence, I think that principle is what's underlying the
"better have a policy, even if that policy is nothing" point in the
existing documents.  But specific nudges in some direction might not
be a bad idea.
  
Alternatively, of course, if there were operations forums where the
details of how exactly to deploy this tricky piece were best
discussed, those forums would presumably also be the best place to
make recommendations on mappings and so on.  I suppose the work going
on at Unicode lies somewhere in that range; there is a whole WG at the
IETF dedicated to DNS operations, and they maybe would have something
to say too.  And, of course, for the TLD and near-TLD space, we have a
large organization half of whose existence is related to assigned
names.  Presumably there are some operators there with some
suggestions on how this all might work.  (In case it's not obvious, to
me the separation of protocol and operational conventions is quite
useful in this particular case, and I think it mightn't be a bad idea
if we were very careful about asking for the protocol to be built
today to accommodate today's peculiar deployment situation.  But I'm
still thinking over the angles before answering the Chair's question.)

A

-- 
Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.


More information about the Idna-update mailing list