[oslc-core] Need for an XML literal value type

Fri Mar 19 13:26:53 EDT 2010

Arthur, thanks for your patience and good answers. I think I'm convinced.

I do have a couple more questions for the WG:

- Should we offer any guidance to folks who wish to store HTML5 or
HTML 4.0.1 content as a property value?

- Instead of an "XML Literal" value-type, what if we allowed only
"XHTML Literal"? -- thus allowing literal XML only for rich text.

Thanks,
- Dave

On Fri, Mar 19, 2010 at 12:30 PM, Arthur Ryman <ryman at ca.ibm.com> wrote:
> Dave,
>
> The data model that OSLC is adopting is RDF. We therefore need to define
> how to represent RDF in various formats. There are some native RDF
> formats, like N3 and Turtle - no problem there. For putting RDF in XML we
> use RDF/XML which is just as much a standard as RDF itself.
>
> In RDF/XML, the way you include XML literal values is via the attribute
> rdf:parseType="Literal" [1]. RDF defines a datatype for literal XML [2].
> This lets us put the angle brackets etc. in the RDF/XML document, and the
> result is a well-formed XML document which is also RDF/XML valid, and we
> can process the literal XML normally.
>
> When we define how to represent RDF as JSON, we need to define how to
> represent the literal values. An XML literal would be encoded in JSON as a
> string, but that string would be valid XML. What problem do you see in
> having a JSON value that contains a valid XML string?
>
> Your proposal to use xsd:string for literal XML would require us to escape
> all the parser-significant  XML characters, which hides the fact that the
> content is XML and defeats processing by standard XML tools (e.g XSLT) and
> would make integration with other RDF data problematic, i.e. how would
> other applications know that xsd:string values needs to be unescaped? How
> would we distinguish strings that were plain text (and might contain angle
> brackets) from text that was really esacped XML?
>
> The background for rich text is that many tools allow users to enter rich
> text and include that in their resource representations. This is
> especially important for Requirements tools where people do spend effort
> to highlight text, e.g. in red, to indicate some semantics. The place
> XHTML comes in is as an interchange format, i.e. OSLC resource should use
> XHTML for rich text. Each tool must convert it's native format to XHTML
> for purposes of interchange via OSLC resources. Uses would not type in
> XHTML directly. They would use editors provided by the tools, and the
> tools would convert it to XHTML for interchange.
>
> In the specific cases of dc:title and dc:description, we should use <span>
> and <div> content respectively. A JSON client that received these literal
> values can simply set or get this as DOM element content  via the
> innerHTML property.
>
> I'd be happy to have a telecon with you to discuss this further.
>
> [1] http://www.w3.org/TR/REC-rdf-syntax/#section-Syntax-XML-literals
> [2]
> http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/#dfn-rdf-XMLLiteral
>
> Regards,
> ___________________________________________________________________________
>
> Arthur Ryman, PhD, DE
>
>
> Chief Architect, Project and Portfolio Management
>
> IBM Software, Rational
>
> Markham, ON, Canada | Office: 905-413-3077, Cell: 416-939-5063
> Twitter | Facebook | YouTube
>
>
>
>
>
>
>
> From:
> Dave <snoopdave at gmail.com>
> To:
> Arthur Ryman/Toronto/IBM at IBMCA
> Date:
> 03/19/2010 09:53 AM
> Subject:
> Re: [oslc-core] Need for an XML literal value type
>
>
>
> On Thu, Mar 18, 2010 at 10:40 PM, Arthur Ryman <ryman at ca.ibm.com> wrote:
>> Please explain why you are concerned. Literal XML is a standard feature
> of RDF.
>
> Thanks for the helpful explanations, Steve, Ian and Arthur.
>
> I guess I still don't understand the need for XML literal in OSLC.
> XHTML can be stored as a string -- so, what are the advantages of
> storing it as literal XML? What specifically does XML literal buy us
> over using a simple string?
>
> I'm concerned here because I'd like to see non-XML representations on
> equal footing and I'd like to encourage folks to express things as
> property values and not as XML constructs. Allowing literal XML could
> open the door to XML content in places other than this XHTML case and
> that could be an issue for JavaScript developers expecting JSON and
> not wanting to ever have to parse XML.
>
> I also had concerns about embedding XML in property values because it
> would result in invalid RDF/XML, but apparently the parse-type literal
> makes this possible. Ian Green's comments seem to contradict this. Do
> we have a definitive answer here?
>
> Thanks,
> - Dave
>
>
>>
>> The pro for XML literal values is that we are using RDF in the way it
> was
>> designed. OSLC data may be combined with RDF data from other sources so
> we
>> should adhere to the standard. We do not want to create an OSLC dialect
> of
>> RDF.
>>
>> For rich text, we should adopt XHTML as the standard interchange format,
>> and we should transfer it in RDF/XML as literal XML, not obfuscate it by
>> turning it into a string.
>>
>> Many development tools capture rich text. Adopting a standard rich text
>> format for OSLC, i.e. XHTML, simplifies processing (e.g. inclusion of
> rich
>> text in UIs, documents and reports) and interchange (so tools only have
> to
>> understand one format as opposed to understand RTF, HTML, etc.). XHTML
> has
>> a simpler syntax than HTML and is XML compliant so it can be readily
>> processed by many XML technologies.
>>
>> In addition, there are other good reasons for using XML in general as
> the
>> value of a property, e.g. when there is an existing XML format, or when
>> plain old XML is a more natural way to represent a literal value
>> (otherwise you get an explosion of blank nodes).
>>
>> Regards,
>>
> ___________________________________________________________________________
>>
>> Arthur Ryman, PhD, DE
>>
>>
>> Chief Architect, Project and Portfolio Management
>>
>> IBM Software, Rational
>>
>> Markham, ON, Canada | Office: 905-413-3077, Cell: 416-939-5063
>> Twitter | Facebook | YouTube
>>
>>
>>
>>
>>
>>
>>
>> From:
>> Dave <snoopdave at gmail.com>
>> To:
>> oslc-core at open-services.net
>> Date:
>> 03/18/2010 09:26 AM
>> Subject:
>> [oslc-core] Need for an XML literal value type
>> Sent by:
>> oslc-core-bounces at open-services.net
>>
>>
>>
>> I'm still a little concerned about adding XML literal as a value type
>> and I'm trying to understand the pros and cons. The only justification
>> that we have so far for adding an XML literal value is for storing
>> XHTML data, which we need for rich text, but we can easily store XHTML
>> data as a string.
>>
>> What specifically do we gain by putting XHTML content in-line in our
>> RDF/XML and Atom XML representations?
>>
>> And conversely, what do we lose by not doing so?
>>
>> Also, does putting XHTML content in-line in RDF/XML result in valid
>> RDF/XML?
>>
>> Thanks,
>> - Dave
>>
>> _______________________________________________
>> Oslc-Core mailing list
>> Oslc-Core at open-services.net
>> http://open-services.net/mailman/listinfo/oslc-core_open-services.net
>>
>>
>>
>>
>
>
>
>