120 likes | 125 Views
Enabling Unambiguous GRDDL Results. David Booth <dbooth@hp.com> Presentation to GRDDL Working Group, 20-June-2007 This document: http://dbooth.org/2007/grddl/ambiguity.ppt. What this issue is about. Given an XML document, what RDF did the GRDDL transformation author intend to denote?
E N D
Enabling Unambiguous GRDDL Results David Booth <dbooth@hp.com>Presentation to GRDDL Working Group, 20-June-2007This document: http://dbooth.org/2007/grddl/ambiguity.ppt
What this issue is about • Given an XML document, what RDF did the GRDDL transformation author intend to denote? • Document receiver needs to be able to determine precisely what graph the XML document was intended to denote • Critical when XML document is a serialization of RDF • GRDDL transformation says how to deserialize
What this issue is NOT about • NOT about desired ambiguity • Transformation author may intend variable results • Can be useful sometimes -- a feature, not a bug • NOT about eliminating ambiguity that the transformation author is willing to accept • NOT about forcing the transformation author to be unambiguous • NOT about requiring the GRDDL-aware agent to actually produce the transformation author's intended results • Users may not need them all, for example • This is only about giving the transformation author the ability to be unambiguous if desired
Kinds of unwanted ambiguity • 1. Wrong results. Results contain different triples than intended. • Source of problem: ambiguous parsing/pre-processing • Proposed solutions: • Proposal 1b part 1, or • Proposal 2b part 1 • 2. Subsetting. User unknowingly gets a proper subset of the intended results, while believing that he/she got all of them. • Source of problem: Lack of a definition of "complete GRDDL results" • Proposed solution: WG Note • Proposal 1b parts 2, 3 & 4
Proposal 1b part 1: Partial solution • Proposal: Change input of GRDDL transformation from XPath node tree to representation • Observations: • Reduces unwanted ambiguity problem • Permits transformation author to reduce variability in the "transformation application" step • Simple normative change • No changes to test cases
Proposal 2b part 1: Better solution • Proposal: • Change input of GRDDL transformation from XPath node tree to representation; and • Specify that parsing is minimal, non-validating/me wn/ • Observations: • Reduces unwanted ambiguity problem • Permits transformation author to reduce variability in the "transformation application" step • Simple normative change • No changes to test cases
Leaving the door open to a WG Note on "Complete GRDDL Results" • Note could define "complete GRDDL results" • Based on existing sec. 7 algorithm • Not required to be implemented • Not on our critical path - could be done anytime before the WG closes
Proposed changes to leave the door open on such a Note • a. Add one sentence to Faithful Renditions. • b. Add non-committal mention of possible WG Note.
XMLDoc RDF XPathnodetree Variability in results -- current spec Transformation Determination Transformation Application GRDDLTransf. Select Representation Parse XSLT/other
XMLDoc RDF XPathnodetree Proposal 1b: Partial solution • Reduces problem in "transformation application" step • But not in "transformation determination" step Transformation Determination Transformation Application GRDDLTransf. Select Representation Parse XSLT/other
XMLDoc RDF XPathnodetree Proposal 2b: Full(?) solution • Limits parsing to minimum, non-validating Transformation Determination Transformation Application GRDDLTransf. Select Representation Parse XSLT/other