Esszett, Final Sigma, ZWJ and ZWNJ

[Paul Hoffman]

At 5:19 PM -0500 2/23/09, John C Klensin wrote:
>I can go back and check (and Cary or Tina may know offhand),

Please do; this is valuable information for the WG in deciding what we should do with the characters that change the bits-on-the-wire names.

> but
>my recollection is that several registries that used variants
>(or other relationships between older and newly-permitted names)
>as part of IDN sunrise policies did just that.   Keep in mind
>that every registry whose core language is written in Latin
>script had to deal (or decide to not deal) with a similar
>problem when it introduced IDNs to expand the "Latin characters"
>beyond ASCII and various conventions for representing decorated
>characters in ASCII to permitting registrations that included
>the decorated characters.

That would make sense, but I don't remember any registry doing that with bundling, and I could not find reference to that in the IDNA WG archives, although I might have missed it.

>And, per Cary's note, sunrise procedures are a plausible policy
>alternative to maintaining variants, as is simply blocking
>registration of the new/ conflicting string.  I prefer variant
>techniques to those alternatives for the same reasons Mark does,
>but our ability to force those techniques on registries is very
>limited -- about the best we can do is to explain the issues and
>hope that they do the right thing.

Fully agree. However, if we have no references to how this worked in the past, we are not in a good position to discuss the pros and cons. This is of particular concern for the secuirty implications that Mark listed.