[Ltru] Re: Preferred-Value cycles

[Frank Ellermann <nobody at xyzzy.claranet.de>]

Doug Ewell wrote:
 
> You are presuming a rule where none exists.

The "dangling pointers are evil" rule is universal.

The one exception ("hakka" misinterpreted as variant)
in 4645bis could be eliminated by registry without
multiple indirections.  It's not allowed to create
another exception under 4646 or similar rules.

> If you are truly "validating," as opposed to checking
> well-formedness, you have to be using the Registry

A registry converter isn't the same as a tag validator,
it just reads stuff top down producing output in a form
needed by other tools, in that case an SGML validator
abused as XML validator.

> you can simply look there to see whether a language 
> "ang" exists and whether an extlang "ang" exists.

Sure, it's possible to check anything directly in the
converter, it could reject extlang "ang" if it already
has language "ang".  And so on, note "Latn" as needed
if seen in a Suppress-Script, and later as "okay" if a
Script "Latn" is registered.

That's possible, and I did that for 4646 registries.
For 4645bis registries I use xml:id for this purpose,
i.e. ID, IDREF, and IDREFS attribute types.

Frank



_______________________________________________
Ltru mailing list
Ltru at ietf.org
https://www1.ietf.org/mailman/listinfo/ltru