Workshop Participants (click on the photo for a larger picture)
See also the separate list of participants.
(Photo courtesy of Alejandro Mallea)
Executive Summary
The Resource Description Framework (RDF), the first layer of the Semantic Web, became a W3C Recommendation in 1999. A major revision was published in 2004, adding a few features, clarifying the syntax and semantics, and retaining compatibility with existing deployment. In the years since then, RDF has been widely implemented and has been adopted in various industries for a wide range of applications. Now, in June 2010, the W3C held a workshop to gather feedback and begin to determine if another revision of RDF is warranted and, if so, which elements should be added or clarified.
The workshop submissions, presentations, and discussions indicated a strong demand for a few features to be added in a compatible manner. Participants also expressed considerable resistance to doing anything which would disrupt or confuse existing deployment efforts. While some participants expressed a strong desire to change certain elements of the design of RDF, there was general agreement that the negative impact from doing so, in nearly all cases, made it unwise.
The work items summary table lists each of the possibilities discussed for an imagined new RDF Working Group. Next to each item is an indication of the level of support it had, expressed via straw poll of the workshop participants. The work items which had strong support and no stated opposition were:
- Adding support for graph identification (such as with named graphs)
- Fixing known errors and shortcomings in the specifications (e.g., using IRI-s instead of RDF URI References)
- Standardizing Turtle
- A syntax related to Turtle to support for embedded/named graphs
- Standardizing an RDF serialization based on JSON
(Note that the workshop participants realized that the term “named graph”, though widely used by the community, is ambiguous, and often refers to what could rather be referred to as quoted graphs, graph literals, etc. It was therefore decided to use the term “graph identification” for the purposes of reporting the workshop’s results, though this term is by no means definitive.)
While the demand was much weaker, there was also wide support and little or no opposition to several more work items, as shown in the table. Based on this data, these items should be left as optional “time permitting” deliverables for a Working Group, allowing them to be addressed if sufficiently motivated participants rapidly develop a design acceptable to the Working Group.
Workshop partipants made it clear that any Working Group chartered should be clearly directed to respect, support, and advance existing deployment, avoiding any changes which would negatively impact current RDF users.
Background
The Resource Description Framework (RDF), the first layer of the Semantic Web, became a W3C Recommendation in 1999. A major revision was published in 2004, clarifying the syntax and semantics, and including separate documents on the general concepts, its semantics, and an XML Serialization (RDF/XML). Since then, RDF has become the core architectural block of the Semantic Web, with a significant deployment in terms of tools and applications.
As a result of the R&D activities and the publication of newer standards like SPARQL, OWL, POWDER, or SKOS, and also due to the large scale deployment and applications, a number of issues regarding RDF came to the fore. Some of those are related to features that are not present in the current version of RDF but which became necessary in practice (e.g., the concept of “Named Graphs”). Others were raised by the design decisions taken in the course of defining the 2004 version of RDF (e.g., restrictions whereby literals cannot appear as subjects). Definition of newer standards have also revealed difficulties when applying the semantics of RDF (e.g., the exact semantics of blank nodes for RIF and OWL, or the missing connection between URI-s and the RDF resources named by those URI-s for POWDER). New serializations formats (eg, Turtle) have gained a significant support by the community, while the complications in RDF/XML syntax have created some difficulties in practice as well as in the acceptance of RDF by a larger Web community.
The RDF Next Steps Workshop was held to consider whether a revision of the 2004 version of RDF is necessary or not. This report constitutes the workshop’s findings.
Workshop Overview
Workshop Format
32 extended abstracts or papers were submitted to the Workshop. The Program Committee accepted 28 of those, these are published on the W3C site. Authors of 18 submissions were also invited to make a presentation at the Workshop (unfortunately, 2 of those could not come to the Workshop after all). There were 35 participants at the Workshop itself (although not everybody could stay for both days). There was a fairly good balance between academia and industry: 18 and 15, respectively (plus two representatives of W3C).
The workshop was conducted over 2 days with a series of presentations consisting of a selection of the accepted submissions. Issues were identified and grouped on a whiteboard and at the end of the first day, based on the submissions and the presentations. Four break-out groups where then formed, which, on the 2nd day, detailed and discussed these issues. The results of the discussion were reported back to the plenary. Discussions during the plenary and some break-out groups were scribed in IRC and the whiteboard discussion was recorded on the Wiki.
Workshop Program Day 1
The first session of the workshop (“Setting the scene”) started with two presentations urging caution with respect to to possible changes by Richard Cyganiak and Jeremy Carroll, but identified a number of issues which they would like to see changed. The 3rd presentation by David Beckett identified a number of issues with current RDF Syntax formats, followed by a presentation by David Booth identifying a number of issues with the current RDF specification.
In the second session of the day a number of specific communities reported on their requirements and issues. Jun Zhao reported on the requirements from the W3C Provenance Incubator Group, Elisa Kendall reported on the experiences the OMG Ontology PSIG has with the current RDF specification. In the final presentation in this session Axel Polleres reported requirements from a number of SPARQL users regarding a combined XML and RDF querying mode.
In the third session a number of presenters reported on specific semantic issues: Alejandro Mallea reported on the mismatch of the current RDF Blank Nodes specification and their usage in SPARQL. Jie Bao highlighted the usefulness of contexts and import facilities for RDF, and Axel Polleres presented the need for annotations of RDF statements. Finally Peter Patel-Schneider presented his motivation to revise the RDF semantics towards a pure data structuring language.
In the last session of the day additional contributions were presented: Fabien Gandon presented the need for named graphs and other graph extensions including RDF/XML syntax extensions. Andy Seaborne highlighted the need for a change propagation mechanism for large RDF graphs. James Leigh presented a RESTful API for isolating changes to RDF data sets and a proposal to make RDF Lists first-class objects. Finally, Ivan Herman initiated a discussion on the usage of owl:sameAs and on what alternatives could be used.
For more details, access to the indvidual presentations’ written submissions, slides, and the discussions, see the workshop program and the workshop minutes.
Collected issues at the end of the 1st day
Issues were collected during the day; these were reordered and categorized at a late afternoon plenary session. Two major categories were “RDF Core” and “RDF Infrastructure”. RDF Core issues were treated as a higher priority for a potential Working Group to consider. RDF Infrastructure issues were elevated for discussion only if they were seen as urgently important for immediate treatment by a Working Group; in general, issues under this category were not felt to be fundamental for a new RDF group, and could be tackled by possibly other groups or the community at large.
For each of these categories the issues were further categorized as “maybe now”, “maybe later”, and “maybe never”. Workshop participants were polled to determine the relative importance of each issue. The most important issues were thus identified by group consensus. Issues considered in the RDF Core “maybe now” category were scheduled for discussion and documentation for the second day.
For each feature there was a lot of discussion on whether these could be introduced without breaking existing systems. In a number of cases, proposals were abandoned simply because it would impact these systems too much. Widespread acceptance and existing deployment (despite there being no standard) played an important role when classifying the results as, say, “maybe now” (Turtle standardization or named graphs are obvious examples).
The results of the first day are summarized in the tables below (see also the photos #1, #2, and #3 taken by Fabien Gandon).
| Identified Issues—RDF Core | |||
|---|---|---|---|
| Issue | Maybe Now | Maybe Later | Maybe Never |
| Graph identification (a.k.a. “Named graphs”) | X | ||
| Turtle | X | ||
| Specifying “follow your nose” | X | ||
| JSON | X | ||
| Atom | X | ||
| XSLT-friendly XML | X | ||
| XML Schema-friendly XML | X | ||
| Skolemize bnodes | X | ||
| Disentangle RDF/RDFS namespaces | X | ||
| URI->IRI | X | ||
| Binary RDF | X | ||
| n-ary predicates | X | ||
| Revise semantics | X | ||
| List construct | X | ||
| RDFa style profiles in, e.g., Turtle | X | ||
| Weakly deprecate some features | X | ||
| Identified Issues—RDF Infrastructure | ||||
|---|---|---|---|---|
| Issue | Maybe Now | Maybe Later | Maybe Never | Discuss on Day 2 |
| Rule-based querying | X | |||
| Identity for non-RDF resources | X | |||
| Evolution (provenance) | X | |||
| Provenance vocabularies | X | |||
| Annotations | X | X | ||
| Standard APIs | X | |||
| Unified RDF/XML/RDB query layer | X | |||
| Change vocabularies | X | |||
| Isolation API | X | |||
| Identity vocabulary | X | X | ||
Workshop Program Day 2
Workshop activities on the second day centered around four broad areas, each assigned to a breakout group. The topics identified at the end of Day 1 for further discussion were distributed among the four breakout groups as follows:
Breakout Group 1: Syntaxes
Chair: Dave Beckett
Topics:
- Turtle
- JSON
- RDFa style Profiles
Breakout Group 2: Semantics
Chair: Paula Gearon
Topics:
- BNodes
- Revising RDF Semantics
- Weak Deprecation
- Literals as subjects
Breakout Group 3: Graph Metadata
Chair: Elisa Kendall See also: IRC minutes
Topics:
- Graph identification (Named Graphs)
- Annotations
Breakout Group 4: Linked Data
Chair: David Booth
Topics:
- Identity Vocabulary
- Follow Your Nose
Results
During the second day, the task forces each drafted a list of work items, with discussion about each item, considering whether and how it might be included in the charter for a new RDF Working Group. There was no time for all workshop participants to review all this text, but it was presented and discussed in the closing plenary session. For more details, see either the 28 June 2010 version of that document (i.e., when the Workshop closed) or the most recent version.
The day (and the Workshop) ended by giving a priority to the core work items discussed during the day. Indeed, the Workshop participants did not endorse all items to be worked on for standardization at this moment. The summary table of results illustrates the consensus of the workshop participants and suggests an order for a later Working Group to address the issues discussed. The issues of most importance, and the ones where consensus was strongest were (in weighted order):
- Standardize Model for Graph Identification
- Modify Semantics to Support Graph Identification
- Fix the inference rules in semantics
- Apply Fixes to Known Spec Errors and shortcomings, including considering postponed issues from the previous RDF Working Group
- Standardize a JSON RDF Syntax
- Make Turtle a W3C Standard
- Add Graphs to Turtle
Recommended Next Steps
The workshop submissions, presentations, and discussions indicated a strong demand for a few features to be added in a compatible manner. Based on the results of the Workshop, and on the community feedbacks thereof, W3C should consider chartering an RDF Working Group at the earliest convenience to address those issues. Workshop participants urge the W3C to use the workshop summary table as guidance in the production of the working group charter. The charter should also list some of the work items of the summary table as “time permitting”, i.e., to be addressed if sufficiently motivated participants rapidly develop a design acceptable to the Working Group.
Acknowledgements
The RDF Next Steps workshop was graciously hosted by Stanford University and the National Center for Biomedical Ontology by providing facilities, refreshments, and lunch for both days of the workshop. Organizers of the Workshop and W3C are grateful to Standford University and NCBO for their hospitality, and especially grateful for the help of Natasha Noy and Chris Macintosh who made all this possible by organizing the workshop logitstics.
The workshop was co-chaired by Ivan Herman, Stefan Decker, and David Wood, assisted by Sandro Hawke. Day 2 breakout sessions were chaired by Dave Beckett, Paula Gearon, Elisa Kendall, and David Booth. Thanks to all the participants for working late and for volunteering their time to improve RDF.