Right. In some situations it might be appropriate for the validator simply
to throw a fatal error; in other situations it might recover by treating it
as a string. It could even interactively ask the use what it should do for
an unknown datatype. I don't think we can dictate behaviour in these
circumstances given the braod range of scenarios in which RELAX NG might be
applied.
--On 01 August 2001 17:05 -0700 Michael Fitzgerald <> wrote:
> I am trying to understand your last statement.
>
> Are you saying that if a RELAX NG schema uses a datatype library that a
> processor does not support, the processor isn't required to validate any
> document against that schema?
>
> -Mike
> -----Original Message-----
> From: James Clark [mailto:]
> Sent: Tuesday, July 31, 2001 4:15 PM
> To: Michael Fitzgerald; RELAX NG List
> Subject: RE: Fallback mechanism for the datatype
>
>
> Saying it may report a warning is vacuous and misleading: a RELAX NG
> processor can give a warning any time it feels like it; it doesn't need
> permission from the spec.
>
> The net result is that if a schema uses a datatype library that a
> validator does not support, then the validator is not required to be able
> to determine whether a document is valid with respect to the schema.
> This is what I think the spec should say.
>
> --On Tuesday, July 31, 2001 10:18 AM -0700 Michael Fitzgerald
> <> wrote:
>
>> I interpret Makoto's paragraph differently. To me it says:
>>
>> "If a conformant RELAX NG processor encounters a datatype library that it
>> does not recognize, it MAY report a warning. It may, for example,
>> continue or abort normal processing."
>>
>> So if I am writing a fridge-firmware implementation, I don't think I am
>> forced to produce a warning or error, and I am happy.
>>
>> -Mike
>>
>>
>> ------------------------------------------------------------------
>> To unsubscribe from this elist send a message with the single word
>> "unsubscribe" in the body to:
>>
>>
>
>
>
>
>
>
> ------------------------------------------------------------------
> To unsubscribe from this elist send a message with the single word
> "unsubscribe" in the body to:
>
>
