xml-dev - RE: A certain difficulty

[Arjun Ray <aray@q2.net>]

On Tue, 22 Feb 2000, Assini, Pasqualino wrote:

> > This led to suboptimal decisions such as redeploying the
> > HTML-inspired idea of sticking a URL (a resource reference) into
> > an attribute.
> 
> Interesting, can you add more on that ?

In RDF, URLs (or "Resources") are first-class data items; in fact, the
assertion chains are all supposed to resolve ultimately to resources,
because "it's all about resources".  There is therefore an imperative
need to *refer* to these resources.  In the SGML/XML formalism, it's
normal to provide for this referential requirement by arranging to
give the "object" a (local) name, and then invoking the name
referentially.  This is what happens with an entity reference, for
instance.  It's also what happens with an IDREF attribute, where the
natural referent is the content of the element with the ID.  The one
thing that's not possible to reference directly is the value of an
attribute.  So, as a rule, you don't put data that you may need to
refer to into attributes - you put them in elements, give the elements
IDs, and use IDREFs to your heart's content.

Yet, RDF does a lot of attribute stuffing with its basic categories of
data.  The basic "reason" for this, of course, is that the normal way
to deal with such data (making it the text content of an element) runs
afoul of -ahem- "what HTML browsers do".  It's called KTWSFN - Keeping
The Web Safe For Netploder.  The result is that the rest of the syntax
has to jump through hoops to reconstruct those basic referential needs
somehow.  I don't even *know* if the syntax actually manages it.


Arjun



***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/threads.html
***************************************************************************