Re: About Refactoring RDF/XML Syntax Rivision 1.43

Satoshi wrote:
> 1. Qname production
My view is that we are now regarding the old grammar as "informal" in that
we are not trying to make sense of what it is defined over (text strings?).
The infoset approach is what I hope will be normative.
In the infoset representation, elements must be namespace qualified. Text to
that effect, maybe with Satoshi's example which are qualified but not
prefixed could be included.
I understand that we have discontinued non-qualified elements. Personally I
think that the modified grammar with the new rule 6.19 captures the
requirement for qualification better than the old rule 6.19; it is then an
XML issue that in fact the prefix can be the default namespace. (Just as the
grammar specifies an order for all attributes, and it is an XML issue that
allows them to be reordered).

> 3. Whitespace handling
>
> It seems that there is no description about whitespace handling. I suggest
> that spec sites that whitespaces contained in elements that never have
> character contents must be ignored.

I like the conciseness of that statement, but suggest it might be too
concise. An example (which I thought I had already posted, but can't find in
the archive) is:

<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
  <rdf:Description>
    <rdf:value>
    </rdf:value>
    <rdf:value rdf:parseType="Resource">
    </rdf:value>
  </rdf:Description>
</rdf:RDF>

The whitespace is significant in the first rdf:value element, and not in the
second.
My attempt
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Aug/0147.html was
"Whitespace is significant in attribute values, where XML whitespace
processing applies, and within literal values. Elsewhere it is ignored."


> 5. Is rdf:RDF optional?
>
> # Aaron has already mentioned about this
>
> Different parsers might generate different models without rdf:RDF,
> especially in case where rdf-ns can be used as a namepsace uri of
> propertyElt. So I suggest that WG recommend to use rdf:RDF for any RDF/XML
> document.

The only cases I've seen which don't have rdf:RDF are standalone files,
hence the rule maybe

RDF ::= <rdf:RDF> description* </rdf:RDF>
     | description

Current M&S does not specify how RDF is signalled when embedded, unless by
rdf:RDF. Does anyone know of an embedding of RDF which does not use rdf:RDF
as the signal to move to rdf processing mode.

Personally I support the proposal to mandate the use of rdf:RDF

> 6. More than one rdf:RDF
>
> Is it legal to have more than one rdf:RDF elements in one XML document?
Yes.
> Is it different from having all rdf:Description in one rdf:RDF?
No.

At least I think so, we probably should make this explicit if that is what
people think. In particular it suggests that using an ID in one rdf:RDF
element in the file, prohibits the use of the same ID in some other rdf:RDF
element.

> 7. Attribute value normalization
>
> Should parsers perform another attribute value normalization for ID or
> bagID, or must not ?


The XML spec is clear.
http://www.w3.org/TR/2000/REC-xml-20001006#AVNormalize
"the XML processor must normalize the attribute value"

Since RDF processing logically follows XML processing we MUST normalize
attribute values.
All we can do is include your example somewhere as a dire warning of things
that shouldn't be done by users who are trying to avoid unnecessary pain.

Jeremy

Received on Thursday, 6 September 2001 09:55:30 UTC