IDNA and U+08A1 and related cases (was: Re: Barry Leiba's Discuss on draft-ietf-json-i-json-05: (with DISCUSS and COMMENT))

[Shawn Steele]

> I'd say that the fundamental design limitation of DNS (never mind IDNA) in this
> context is that a domainname's labels are supposed to make good identifiers
> (observe: hand-waving about what an "identifier" is).

I think this is a big part of the problem.  In some contexts an identifier has to have a 1:1 mapping with an idea and round trip.  So if I have a "thing", then I can figure out what its one (and only) identifier is.  

In other cases though, we want identifiers to be a little fuzzy.  It's convenient that DNS maps upper & lower case ASCII (unless you're Turkish) for example.  Software often lowercases them so it ends up with it's 1:1 mapping.  So you end up with an n:1 system that will probably spit out a 1:1 label after it's been processed.  It worked mostly because the casing rules were consistent and used everywhere.

Not exactly a confusable, but I'd've mapped ALL 4 of the Latin I forms together.  Then there's no confusion about which identifier is used for a label but it wouldn't always round-trip to a "pretty" label.  Some have made an argument that DNS doesn't have to be pretty though, and if people wanted to advertise the pretty form it'd still resolve to the same thing.

-Shawn