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

Andrew Sullivan ajs at shinkuro.com
Tue Dec 1 19:04:38 CET 2009

On Tue, Dec 01, 2009 at 05:43:41PM +0000, Shawn Steele wrote:
> A) If nothing else the English words at least have consistent
> mapping rules for casing to apply, how does mapping lowercase SS to
> the "correct" word?

No, there are not mapping rules at all for ASCII characters.  The DNS
is case-preserving but case-insensitive, which is slightly different
from mapping.  What's worse, not every implementation follows these
rules precisely, and in any case the rules are extremely loosely
specified and don't work properly with compression pointers.  The
upshot is that the DNS doesn't actually work today the way you keep
claiming it does: the "mapping rules" simply don't work the way you
insist they do, and that's why the mixed-case tricks that were
envisioned as the primary benefit of the case preservation don't
actually work in practice.  The upshot is that users have a rough idea
of how things will work, and they use that rough idea in practice.
Their success reinforces their ability, and that's how they end up
"knowing" how this works.  Since every other technical thing that
people learn is also learned this way (we learn by doing, as people
have observed since at least Aristotle), it is reasonable to imagine
that people will adapt for the case of non-ASCII characters too.

> B) The casing alone of eszett would seem to indicate some linguistic
> equivilance of eszett ~ SS that doesn't exist for o ~ OU.

I don't think I understand what that means. 

> C) In addition it's the existing practice of IDNA2003, and there's not really good rules.

Right.  We chartered work to make changes to IDNA2003.  You seem to be
arguing that we should not have done that.  As I already remarked,
that's a big claim and I think you will need to defend it.

> D) and lastly I really doubt you will find a German speaker that
> wanted to protect their trademark that would think it was "safe" to
> let a potential competitor use the other form of their legally
> protected name.  If you can get someone from the legal dept of
> fussball.de to say that the won't sue if some one else takes
> fußball.de then I'll throw in my virtual towel.

That's a total red herring.  We have this problem today, with
trademarks that are easily confusable under plain-ASCII, no-bundling
rules, and everyone has figured out what to do about it.  Your
argument depends on a false dichotomy: either we put the restriction
in the protocol, or else all manner of dirty nasty subversion of the
international trademark situation will happen.  That's preposterous on
the face of it: the only thing controlling domain name registration is
certainly not the DNS rules, and yet we have managed to build a layer
of policy enforcement so that the kinds of problems you're talking
about don't happen without warping the DNS so that colourmyworld and
colormyworld are automatically the same label.

If your argument comes down to, "I don't trust registry operators --
i.e. zone operators -- to do the right thing, so I want everything
sewn up in the lookup protocol," then I think you're out of luck.  Not
only did that ship sail more than 20 years ago, but I don't believe
you could build a protocol that would meet the need anyway.  Finally,
and most importantly, if you can't trust the operators of the zones
you have much bigger problems anyway, because the fact is that you
have to rely on the data they provide, or you can't look anything up.


Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.

More information about the Idna-update mailing list