Lists Home |
Date Index |
The real problem with XPointer is that we *really* need a simple way to
point into XML documents. The more specialized functionality of XPointer is
also useful, but to a much smaller audience. Unfortunately, XPointer stands
in the way of simple XML pointers, because XPointer owns the fragment
identifier. Consider the following excerpt from the XPointer abstract:
>This specification defines the XML Pointer Language (XPointer), the
>language to be used as the basis for a fragment identifier for any URI
>reference that locates a resource whose Internet media type is one of
>text/xml, application/xml, text/xml-external-parsed-entity, or
In HTML, most pointers are done with very simple pointers like this:
If XPointer owns the fragment identifier, any implementation that supports
the above form of pointers must support a superset of XPath in all its
glory. But pointers should be simple and fast. Many applications need to
quickly traverse pointers among XML documents, but have no need to
implement XPath to do so. For instance, a query language needs to be able
to traverse references among documents, but can optimize much better if
these references are not written in their own query language.
So I believe that XPointer has a right to exist, but I don't think that the
full-fledged XPointer should own the fragment identifier. It must be
possible to have pointers among documents without implementing all of XPointer.
>At 12:36 PM 2/1/2002 -0500, Daniel Veillard wrote:
> And XPointer is small compared to other specifications, quite smaller
>than XSLT which reuses XPath too, not that much longer than for example
>RFC 2396, but sure it builds on top of more existing layers. But in itself
>it is relatively small.
XPointer is a way to point to a location in an XML document.
XSLT is a full-featured transformation language.
I would *expect* XPointer to be smaller than XSLT. I would expect it to be
a *lot* smaller, especially since XPath already allows you to point to any
node in an XML document.
I think XPointer is useful, but it should not own the fragment identifier.
That should be reserved for something *much* simpler.