RE: Fwd: I18N Last call comments on Schema Part 2

>I'm sorry, but it seems that your comments are private comments,
>but that I misunderstood them to be comments from the WG.

I'm not a W3C member, just an interested outsider.  If I stated any opinions about the WG views or intentions, it was based on observations and the Psychic Hotline (press 1 for Romance, press 2 for
Career, press 3 for XML Schema WG).

> If the WG is not really interested, it's probably lost time.

True.  Many of the points that you raised have been discussed before and the schema WG has not shown any indication of pursuing them.  If the I18N group is going to raise the issues, then it would be
good to know the context of previous discussions.

>That's easy, but how do you do Islamic or Hebrew calendars?

Short of putting a full scripting language or access to external translators in Java , both of which are undesirable, you are not going to be able to do every translation that you ever might want to
do.  Regex replacement and translate might get you 95% or more with minimal additional complexity.

It might be possible to indicate that something has a the value space of a particular type but the transformation is not possible to express using the capabilities of schema.

<xsd:simpleType name="hebrewDate" base="xsdt:date">
	<xsd:transform external="true">
		<xsd:annotation>
			<xsd:appinfo>
				<object class="com.example.HebrewDateTransform"/>
			</xsd:appinfo>
		</xsd:annotation>
	</xsd:transform>
	<!--  this constraint could not be enforced by a parser   -->
	<xsd:maxInclusive value="2002-12-31"/>
</xsd:simpleType>

>I think you misunderstood some of our comments. We do not want one
>single lexical representation for all numeric datatypes. We want a single
>lexical representation for all types derived from the same base type,
>i.e. if something is float or double, it has an E. If something is
>decimal or derived thereoff, it doesn't have an E.

My interpretation of you comments was an attempt to have a canonical representation on any particular value, that "1E2" would be the one and only legal lexical representation and "100E0", "1.0E2",
"10E1", "+1E2", etc would not be legal.  This seems to be more and issue of canonicalization than internationalization.  Schema appears to give you sufficient functionality to write a canonical schema
(a pattern could restrict doubles to just canonical forms), however forcing datatypes to be canonical by itself is not sufficient to make an arbitrary schema canonical and would be a significant
irritation to those who aren't concerned about canonicalization.

Received on Friday, 2 June 2000 11:29:34 UTC