Stability of valid IDN labels

[John C Klensin]

Mark,

For better or worse, Eric is probably right.   Because some 
implementations of applications make tests against the names of 
TLDs and then decide, on that basis, how to proceed, there is 
less difference between a rule about the syntax of a domain name 
and particular registration status than one might imagine in 
another environment.  Those tests are bad practice in some cases 
and essential in others, sometimes both.

More generally...

> Note: the migration from CONTEXT to PVALID is effectively
> already allowed in the drafts; it just means loosening the
> context rules until they have no effect.

This was quite intentional.  However, the reason for the 
distinction between "context for joiners" and "other context" 
had to do with checking at resolution time.  If we preserve the 
principle that there is a resolution-time check so that putative 
labels containing DISALLOWED characters are not looked up, then 
a character that is moved from DISALLOWED to CONTEXT (or PVALID) 
won't be accessible to any software that has not been updated. 
I think that maybe that is ok, or at least less bad than the 
alternatives, but I don't see how and whether it makes a 
difference whether a character is moved from DISALLOWED to 
CONTEXT rather than DISALLOWED to PVALID.

1) allow characters to migrate from DISALLOWED to CONTEXT.

See above.

2) disallow CONTEXT rules from becoming more restrictive.

I think this is ok and desirable.

In a sense, the real stability criterion one would like is "if 
this was valid at the time it was registered, then it is valid 
forever".  Unfortunately, that criterion is impossible to state, 
much less enforce, given the many millions of zones in the DNS.

     john