RE: [xml-dev] Brain Teaser: Element Author is of type xsd:string, what's

["Costello, Roger L." <costello@mitre.org>]

Thanks Michael and Noah.

I would like to summarize.


1. This is the set of legal characters in XML 1.0

#x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]
/* excluding the surrogate blocks, FFFE, and FFFF. */


2. This is the set of legal characters in XML 1.1

[#x1-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]  /* excluding the
surrogate blocks, FFFE, and FFFF. */


3. These are the characters which are not legal, regardless of whether
XML 1.0 or XML 1.1 is used:

#x0  [#xD800-#DFFF] [#xFFFF-#xFFFFF] [#x110000-#xFFFFFF] #xFFFE #xFFFF


4. Suppose element Author is declared to be of type xsd:string.  In
each example below the Author element has invalid content:

    <Author>&#x0;</Author>
    <Author>&#xFFFE;</Author>
    <Author>&#xFFFF0;</Author>
    <Author>&#xFFFF1;</Author>


5. I tested this with XML Spy and Oxygen XML.  Here are the results:
 
                                  XML Spy         Oxygen XML
---------------------------------------------------------------
    <Author>&#x0;</Author>         Valid            Invalid
    <Author>&#xFFFE;</Author>      Valid            Invalid
    <Author>&#xFFFF0;</Author>     Valid            Valid
    <Author>&#xFFFF1;</Author>     Valid            Valid

XML Spy incorrectly validates the data in all four cases.

Oxygen XML correctly validates the data in two cases, and incorrectly
validates the data in the other two cases.

Is the above summary correct?

/Roger



-----Original Message-----
From: noah_mendelsohn@us.ibm.com [mailto:noah_mendelsohn@us.ibm.com] 
Sent: Friday, February 09, 2007 10:48 AM
To: Michael Kay
Cc: Costello, Roger L.; xml-dev@lists.xml.org
Subject: RE: [xml-dev] Brain Teaser: Element Author is of type
xsd:string, what's an illegal value of Author?

Michael Kay writes:


> There's a W3C note on handling XML 1.1 with XML Schema:
> 
> http://www.w3.org/TR/xml11schema10/
> 
> and it recommends (see the last line of the note) that with that
> combination, the definitions of the built-in types should be
"stretched" 
to
> accommodate the characters allowed in XML 1.1. With that strategy,
there
> will never be an invalid instance of xs:string.

Indeed.  That's about the best we could do without appearing to make a 
retoactive incompatible change to Schema 1.0.  With Schema 1.1, such 
incompatibilities are less of an issue.  Note that the latest public 
working draft of Schema 1.1 says [1]:

"[XML Schema: Datatypes] defines some datatypes which depend on 
definitions in [XML 1.1] and [XML-Namespaces 1.1]; those definitions,
and 
therefore the datatypes based on them, vary between version 1.0 ([XML 
1.0], [XML-Namespaces 1.0]) and version 1.1 ([XML 1.1], [XML-Namespaces

1.1]) of those specifications. In any given
schema-validity-·assessment· 
episode, the choice of the 1.0 or the 1.1 definition of those datatypes
is 
implementation-defined."

The working draft for Schema 1.1 Datatypes provides more details [2]:

"This specification defines some datatypes which depend on definitions
in 
[XML] and [Namespaces in XML]; those definitions, and therefore the 
datatypes based on them, vary between version 1.0 ([XML 1.0],
[Namespaces 
in XML 1.0]) and version 1.1 ([XML], [Namespaces in XML]) of those 
specifications. In any given use of this specification, the choice of
the 
1.0 or the 1.1 definition of those datatypes is implementation-defined.

"Conforming implementations of this specification may provide either
the 
1.1-based datatypes or the 1.0-based datatypes, or both. If both are 
supported, the choice of which datatypes to use in a particular
assessment 
episode should be under user control.
Note:  When this specification is used to check the datatype validity
of 
XML input, implementations may provide the heuristic of using the 1.1 
datatypes if the input is labeled as XML 1.1, and using the 1.0
datatypes 
if the input is labeled 1.0, but this heuristic should be subject to 
override by users, to support cases where users wish to accept XML 1.1 
input but validate it using the 1.0 datatypes, or accept XML 1.0 input
and 
validate it using the 1.1 datatypes. "

Regarding "string" in particular, the Datatypes draft says [3]:

"It is implementation-defined whether an implementation of this 
specification supports the Char production from [XML], or that from
[XML 
1.0], or both. See Dependencies on Other Specifications (§1.3)."

So I think we've been quite careful with the details there, and Schema
1.1 
can indeed be used to validate as strings the characters that Roger 
mentions.  Note that there are some characters that are not allowed in 
either XML 1.0 or XML 1.1 infosets, and thus are disallowed even in
Schema 
1.1 strings.  I believe I'm correct that NUL (0x0) is one of these.

Noah

[1] http://www.w3.org/TR/xmlschema11-1/#intro1.1
[2] http://www.w3.org/TR/xmlschema11-2/#intro-relatedWork
[3] http://www.w3.org/TR/xmlschema11-2/#string
--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------