Re: RDF Frequently Asked Questions: input needed

[Here are some of the most common questions I get asked about RDF]

Q: Can I use a DTD with RDF (or, with a more hostile tone, WHY ISN'T there a
DTD for RDF)?

A: You can certainly define a trivial DTD for the elements and attributes in
the RDF namespace, but if your objective is to use parser enforced
validation, an XML version 1.0 DTD won't cut it.  RDF uses attribute and
element namespace expanded names as data, which makes traditional validation
impractical.  Higher-level validation is possible through RDF Schema and
domain/schema specific validators.



Q: Should I use RDF or XML?

A: Yes.  :-)

Seriously, this is a difficult question, and it isn't clear that there is a
a concensus answer.  The process of deciding whether an XML format that
conforms to RDF is preferrable to one that doesn't involves a lot of
criteria, constraints, and a certain amount of guesswork.

Here are the things for which I think RDF XML is superior to non-RDF XML:
	* Semantic Web
	* Simple or legacy key/value metadata
	* Metadata that requires structured containers or sub-properties
	* Second-order metadata (statements about statements)

Here are the things for which I think non-RDF XML, particularly when
controlled by XML Schema, is superior to RDF XML:
	* Interchange of RDBMS table data
	* Structured content data (as opposed to metadata)
	* "High performance" metadata (where time or space is an issue)



Q: Can a property value be inline binary (like a thumbnail image)?

A: The RDF syntax is based on XML, and inherits all the same advantages and
disadvantage of XML.  While inline binary is possible, it isn't pretty. You
can always use an URL reference as a property value, but assuming that isn't
going to work for you, your only other option is to filter the binary into a
safe representation, whether through a CDATA section, or through character
escapes, or through an equivalent encoding, such as base64.



Q: Why is the RDFMS specification so terrible/complicated?

A: Personally, I think it is one of the better written specifications. 
Unlike a lot of nuts&bolts specs, it actually provides some introductory
material, and motivates at least part of the design with actual examples.

Still, many smart people read the spec three or four times and still have a
hard time understanding it.  [It would be worthwhile digging into this a
little to find out why -- I never did myself].

I think that one contributing factor might be that the XML community at
large isn't quite ready to believe that XML can be used to represent
anything but an attribute decorated tree.  The directed named-arc graphs of
the RDF data model confounds those who are expecting to see a tree. 
Furthermore, I think there is a lot more attachment to XML syntax than the
RDFMS specifiers realized.  I find that a lot of people want to infer the
model from the syntax, when learning a new XML application.  You can look at
an example of SVG or SMIL and get a general idea of what is going on.  It's
much harder to do this from an example of RDF.  Of course, this inference
and intuition is unnecessary, since the RDFMS spells out the model FIRST, so
go figure.

Perry

Received on Tuesday, 4 April 2000 19:55:23 UTC