Re: ANN: VoiceXML in RELAX NG

[James Clark <jjc@jclark.com>]

I noticed you translated an attribute like:

  version CDATA #REQUIRED

into

  <attribute name="version">
    <data type="normalizedString"/>
  </attribute>

I would suggest it is better to use one of:

  <attribute name="version"/>
  <attribute name="version">
    <data type="string"/>
  </attribute>
  <attribute name="version">
    <text/>
  </attribute>

Firstly, an XML parser always does CDATA-normalization of attribute values, 
even if it's non-validating and even if there is no DTD.  Thus, if the 
schema declares the type simply as string, the schema normalized value with 
or without the DTD being parsed will always be the same as the normalized 
value that would have been reported if the DTD had been parsed.

Secondly, the normalization done by normalizedString is not the same as 
CDATA-normalization.  Specifically, CDATA-normalization does not affect 
characters specified in numeric character references.  So, for example, if 
I had an attribute:

  version="&#xA;"

then the normalized value with the DTD would contain a newline, but the 
schema normalized value with <data type="normalizedString"/> would not 
contain a newline.

On a separate point NMTOKENS is NMTOKEN+ not NMTOKEN*, so you need to use 
list/oneOrMore not list/zeroOrMore.

You might want to try DTDinst, instead of converting by hand. It can do a 
very good job on a DTD like this.

--On 05 September 2001 14:13 -0700 Kohsuke KAWAGUCHI 
<kohsuke.kawaguchi@eng.sun.com> wrote:

>
> I wrote a VoiceXML in RELAX NG. It is available at:
>
> http://www.geocities.com/kohsukekawaguchi/relaxng/voicexml.html
>
>
> regards,
> --
> Kohsuke KAWAGUCHI                          +1 650 786 0721
> Sun Microsystems                   kohsuke.kawaguchi@sun.com
>
>
> ----------------------------------------------------------------
> To subscribe or unsubscribe from this elist use the subscription
> manager: <http://lists.oasis-open.org/ob/adm.pl>
>
>