LC Responses/FH3

From OWL
Jump to: navigation, search

Shorter reply as discussed at F2F:


To: Frank van Harmelen <Frank.van.Harmelen@cs.vu.nl>
CC: public-owl-comments@w3.org
Subject: [LC response] To Frank van Harmelen

Dear Frank,

Thank you for your comment
     <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0036.html>
on the OWL 2 Web Ontology Language last call drafts.

We would like to stress that RDF/XML is the only syntax that MUST be supported by implementations; support for the XML syntax is not required. We have revised several of the documents and introduced a new Document Overview [3] in order to make this clearer.

The XML syntax is motivated by the desire to support OWL users who want better interoperability with XML based tools and languages, for example WSDL, XSLT/XQuery/XPath, or schema aware editors. An additional benefit is that XML data can be exposed to RDF/OWL applications using GRDDL (see [1]). We agree that it is worthwhile to mention this in New Features and Rationale [2], and the current version of the document has already been improved in this respect.

[1] http://www.w3.org/2008/06/wiki_scribe/?source=http://www.w3.org/2007/OWL/wiki/Chatlog_2009-02-24#resolution_16
[2] http://www.w3.org/2007/OWL/wiki/New_Features_and_Rationale
[3] http://www.w3.org/2007/OWL/wiki/Document_Overview

Please acknowledge receipt of this email to <mailto:public-owl-comments@w3.org> (replying to this email should suffice). In your acknowledgment please let us know whether or not you are satisfied with the working group's response to your comment.

Regards,
Ian Horrocks
on behalf of the W3C OWL Working Group



CUT AND PASTE THE BODY OF THE MESSAGE (I.E. FROM "Dear" TO "Group") INTO THE BODY OF AN EMAIL MESSAGE. SET THE To:, CC:, AND Subject: LINES ACCORDINGLY.

PLEASE TRY TO REPLY IN A WAY THAT WILL ALLOW THREADING TO WORK APPROPRIATELY, I.E., SO THAT YOUR REPLY CONTINUES THE THREAD STARTED BY THE ORIGINAL COMMENT EMAIL



We are positively opposed to introducing the (non-RDF) XML

serialisation. It's entirely up to the WG to adopt a different structural specification / abstract syntax if that makes it easier to define the semantics of the language etc, but we can see no good reason for turning this internal formalisation-tool into a normative additional serialisation syntax. The merits of the new syntax are justified in terms of debatable arguments (a preference for either axiomatic or frame-based syntax is a matter of taste, not fact). We also don't see how the introduction of two serialisation syntaxes (RDF-XML and non-RDF-XML) can make life easier for developers. Arguably the non-RDF-XML syntax is easier to handle, but it is mandatory for tools to implement the RDF-XML syntax, so it's just additional burden. (see conformance statement 2.1 in <http://www.w3.org/TR/owl2-test/>) Also, introducing the non-RDF-XML syntax breaks upwards compatibility: Without any modification, many OWL1 tools will be able to parse all of OWL2 in the RDF-XML syntax and most likely even make some semantic sense of it, while the same OWL1 tools will barf at (or at best entirely ignore) the same OWL2 ontologies when expressed in the non-RDF-XML syntax. It is also noticeable that the Features document does not give any supporting use-cases for the introduction of the new syntax. Summarising: this will be a burden on tool developers, and will break compatibility. Finally, it breaks with the widespread semantic-web practice that triples are the exchange currency. In short: we strongly insist that this syntax will be made non-normative (comparable to the non-RDF-XML syntax for OWL1)

In relation to this previous point, we would find it useful if the documents would show the triple-serialisation (ie in RDF-XML) of the various new constructs, similar to the Guide document for OWL1 <http://www.w3.org/TR/owl-guide/>

Frank van Harmelen, and many members of the Semantic Web Group at the Vrije Universiteit Amsterdam


Start of a reply

From: Bijan Parsia <bparsia@cs.man.ac.uk> Date: Mon, 26 Jan 2009 18:56:07 +0000 Message-Id: <9926856B-8AF7-4F74-89DC-6C3AEE607EC9@cs.man.ac.uk> To: W3C OWL Working Group <public-owl-wg@w3.org>


http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/ 0036.html

My responses:

Please note that the XML syntax is not primarily justified in terms of its relation to the functional syntax. Rather it is independently motivated and follows the functional syntax style in order to simplify use, learning, specification, etc. In other words, if one is going to have an XML syntax, there needs to be special reason to

  • depart* from the functional syntax. There is no such reason.

The motivation for the XML Syntax is better integration with the huge XML infrastructure, toolchain, and better accessibility for the XML savvy user base. For example, RDF/XML, practically speaking, is not XML Schema-able. Thus, it's difficult, or impossible!, to use OWL in WSDL based web services in a type sensible way. For example: <http://www.w3.org/mid/200710311232.31360.matthew.pocock@ncl.ac.uk>

With a Schema-able serialization, it is possible and practical to use generic XML tools. I, for example, use oXygen heavily and get things like auto-completion for free.

Writing useful XPath, XSLT, XQuery and CSS is prohibitively difficult for RDF/XML, while straightforward for OWL/XML. It is impossible, afaik, to do schema aware queries over RDF/XML (i.e., query for all axioms, or all class axioms, without having to do a union query), but straightforward against OWL/XML.

Many organizations are heavily invested in XML tooling and training, thus it behooves us to do reasonable outreach.

"""We also don't see how the introduction of two serialisation syntaxes (RDF-XML and non-RDF-XML) can make life easier for developers. Arguably the non-RDF-XML syntax is easier to handle, but it is mandatory for tools to implement the RDF-XML syntax, so it's just additional burden."""

OWL/XML actually reduces the burden of development for a large class of developers --- those of XML centric tools. So, we have to be a little careful about how we measure burden.

Secondly, there's already an open source toolkit (the OWL API) and web service: <http://owl.cs.manchester.ac.uk/converter/>


"""Also, introducing the non-RDF-XML syntax breaks upwards compatibility: Without any modification, many OWL1 tools will be able to parse all of OWL2 in the RDF-XML syntax and most likely even make some semantic sense of it, while the same OWL1 tools will barf at (or at best entirely ignore) the same OWL2 ontologies when expressed in the non-RDF-XML syntax."""

Given the OWL API and web based converter, it's not difficult for developers to add it in or users to do the conversion themselves (after all, this is the case with Turtle right now). We expect further translators (e.g., XSLT or XQuery based) to emerge during CR. Finally, if there is significant use of it, we imagine major of OWL tools will be happy to add a fairly easy to parse format to their toolkit in exchange for an expanded user base.

"""It is also noticeable that the Features document does not give any supporting use-cases for the introduction of the new syntax."""

The features document is not in last call and is not complete. We shall add the rationales listed above.

"""Summarising: this will be a burden on tool developers, and will break compatibility."""

We believe that the burden is overstated and highly mitigated by the available of open source converters. The benefit to users and application developers is very high as is the potential to expand the OWL market and generally reach out to the XML community.

"""Finally, it breaks with the widespread semantic-web practice that triples are the exchange currency."""

First, as OWL/XML has a precise, well speced, and standardized translation to triples, it's clear that OWL/XML counts as triples. It does not count as RDF/XML, but then, neither does Turtle. Turtle is a syntax that appeals to people authoring by hand without tool support. OWL/XML appeals to XML people. These are good rationales for having these additional syntaxes.

Furthermore, with GRDDL and with RIF (which has only an XML exchange syntax, afaik, with no RDF mapping) it seems that this semantic-web practice is not a trump. We can depart from it for good reason and we have, in this case, good reason.

Finally, it does not make sense to make it non-normative. If we are trying to spec things for the XML toolchain, then we should do so normatively. Neither RDF/XML nor the RDF triple model satisfy this need. Note, that this doesn't displace RDF, but it does better connect OWL with the XML world. That's good for the semantic web.

Cheers, Bijan.


PeterPatel-Schneider 18:47, 21 February 2009 (UTC) A reply that I put together, not noticing that Bijan had one already. Note the bit at the end about triples.

[Draft Response for LC Comment 32:] CO1

Dear Frank,

Thank you for your message

 http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0036.html

on the OWL 2 Web Ontology Language last call drafts.

The Working Group felt that there would be adequate interest in a native XML syntax for OWL 2 to support the development of such syntax, which would work well with XML tools. For example, XQuery tools can extract certain kinds of syntactic information out of OWL ontologies written in the XML syntax. The working group felt that it was the lack of this recommendation status that limited the use of the native XML syntax for OWL 1, and thus the native XML syntax for OWL 2 is a recommendation-track document.

The particular XML syntax for OWL was chosen to match closely the structural form of OWL 2 constructs. This was a conscious decision, to support ease of use for those tools that manipulate OWL 2 ontologies in the structural form.

As you note, RDF/XML is the primary exchange syntax for OWL 2. Any OWL 2 tool that stores OWL 2 ontologies and makes them available in the Semantic Web is required to use RDF/XML as its exchange syntax if asked to do do (unless the ontology cannot be converted to RDF/XML, of course). Therefore no tools are required to implement *any* OWL 2 syntax other than RDF/XML, regardless of the recommendation status of the native XML syntax for OWL 2.

OWL 1 tools will therefore be able to receive (almost) any OWL 2 ontology in RDF/XML and may be able to perform useful actions on them. Note, however, that unmodified OWL 1 tools may not produce completely correct answers when processing OWL 2 ontologies exchanged using RDF/XML, just as an RDF-only tool may not produce completely correct answers when processing an OWL 1 ontology exchanged using RDF/XML, or even when processing an RDFS document exchanges using RDF/XML. The recommendation status of the native XML syntax for OWL 2 does not affect this situation in any way.

Tools where the update of the native XML syntax for OWL 2 is expected generally fall in the category of tools that need significant changes to correctly handle OWL 2. As you note, there is an additional burden on these tools to support the native XML syntax for OWL 2, but initial results indicate that builders are willing to take on this additional burden, and indeed some have already done so.

The Features document is not yet finished, and has concentrated on the additional expressivity in OWL 2. It is expected that the Features document will eventually record the rationale for all the new features of OWL 2.

In summary, support of the native XML syntax for OWL 2 is only a burden for those tools and developers who choose to support it, all other tools and developers do not need to do anything in response to its recommendation status; no compatibility issues are raised because of its recommendation status; and RDF/XML is primary exchange syntax for OWL 2.

Therefore the working group does not intend to change the status of the native XML syntax for OWL in response to this part of your comments.


There are several documents that provide triple serializations of the new constructs in OWL 2. The Primer <http://www.w3.org/TR/owl2-primer/> has examples of the new constructs of OWL 2 in RDF/XML and, indeed, is designed to cover most of the needs previously met by the OWL 1 Guide. There is also an as-yet-unpublished Quick Reference Guide for OWL 2, currently at <http://www.w3.org/2007/OWL/wiki/Quick_Reference_Guide>, that provides triple serializations for each construct, both old and new. The normative specification of the triple serializations of all constructs in OWL 2 is the OWL 2 Mapping to RDF <http://www.w3.org/TR/owl2-mapping-to-rdf/>, one of the OWL 2 documents currently in last call.


Please acknowledge receipt of this email to <mailto:public-owl-comments@w3.org> (replying to this email should suffice). In your acknowledgment please let us know whether or not you are satisfied with the working group's response to your comment.

Regards, Peter F. Patel-Schneider on behalf of the W3C OWL Working Group