Re: RDF/XML (was Re: My action item on RDDL/RDF)

Perhaps I'm naive but I don't think this thread will go on long so I 
prefer to finish it up here.

Dave Beckett wrote:
> ...
> 
>>Second, this is just further evidence that RDF/XML is broken. 
> 
> I saw no specific evidence, just things like "RDF tax" and "doomed"
> which are not very informative.

You've already said that you  know the problems. Everybody else already 
knows the problems. Re-iterating them isn't a useful expenditure of my 
time or Tim's. If you know that there are huge problems and the only 
thing preventing you from fixing them is process then I have no idea why 
you take offense to the assertion that the current syntax is broken.

>...
>>I don't see anything in the grammar that allows you to put string values 
>>in attribute values rather than content so it is a huge failure if 
>>legacy browser support is a goal.
> 
> 
> You missed it reading the grammar.   Look up property attributes.

Yes, I don't know why I write that. I had used property attributes 
earlier that day. I guess I was thinking about mixed content.

>...
> You can have literal XML content (anything you like, mixed if you
> want) as RDF/XML statement values.  How you define easy is
> subjective.

Simple:

<Class>
    <Property>
       <rdf:literal>
          <p>This <emph>is</emph> easy.</p>
       </rdf:literal>
     </Property>
</Class>

> This paragraph has three !s so seems rather excited.

It is.

> If I try to pull out your points:
> 
>   1.  No easy way to do literal XML as value in RDF/XML
>   2.  No easy way to do mixed-content as value in RDF/XML
>   3.  No easy way to do other type-declared data in RDF/XML
>   4.  "it" (not clear what) must be fixed in RDF [ or do you mean
>       RDF/XML the transfer syntax? ]

At least the transfer syntax. There should probably be thought given to 
whether it is desirable to have some form of use-by-reference of the 
infoset in the RDF model and vice versa. But fixing the RDF/XML syntax 
is a good start.

>   5.  RDF's XML syntax is hostile to "normal ways of using XML"
>   6.  Any rational integration of XHTML and RDF has to address XHTML inside RDF
> 
> 1-3: "easy" is subjective.   1&2 are supported in RDF/XML so I don't
>   see what needs to be "fixed" - If you have specific problems, I don't
>   see them.  Given it is possible to do 1&2, it just comes to if if meets
>   your definition of "easy".  Somehow I doubt it.

I see a syntax for RDF to refer to mixed content and literal XML. I see 
nothing for having RDF embed mixed content and literal XML.

> 3: I don't understand "other type-declared data" - please cite
>    something more specific.

<Class>
    <Property>
       <rdf:literal rdf:type="xsi:integer">5</rdf:literal>
     </Property>
</Class>

<Class>
    <Property>
       <rdf:literal rdf:type="xsi:real">5</rdf:literal>
     </Property>
</Class>


> 4: Still can't see your requirement.
> 5: Normal XML?!  Ha ha.  RDF/XML might look old since it was designed
>   around 1997/1998 at the same time as XML and XML Namespaces.  I guess.

First, it is really irrelevant to what needs to be done today how this 
situation came about.

Second, there was more than 20 years of SGML experience to draw upon at 
the time. Anybody with SGML experience should have been able to see that 
mixed-content literal values would be important.

> 6: Wouldn't that be up to XHTML to decide how to embed other
>    languages?  <object>?  I haven't looked at what they are thinking of.

According to your transliteration of my post, 6. is:

"Any rational integration of XHTML and RDF has to address XHTML inside RDF"

XHTML addresses RDF in XHTML. The RDF core team has to address XHTML 
(and other vocabularies that use mixed content) in RDF.

> I'd hope this is in the vein of gathering requirements for new
> syntaxes, rather than chopping and changing this one.

I would say that some of the problems of the current syntax could be 
fixed. A richer vocabulary of literals could be a new feature. Having an 
entirely new syntax is probably also important as another project.

> ...
> Yes.  That's the striping.  We know.
> ...
 > This has been suggested many times.  We all know wouldn't start from
 > here for an RDF/XML syntax in 2002 versus 1997/1998.

My point exactly. Everybody knows the problems already. Given the 
experience of the last several years we now know that the syntax is 
_broken_ as in extremely difficult to use.

>....
>>... IMHO, there should be a first class XML->RDF mapping language.
> 
> 
> Interesting.  I think there has been some work on that already - I'd
> have to look up the pointers, maybe Uche Ogbuji's work in Versa?

You're right that Uche Ogbuji has done work on it, but not in Versa, 
that's the query language.

  Paul Prescod

Received on Tuesday, 12 November 2002 09:48:16 UTC