proposed response to LC comment 7

[Response for LC Comment 7]

Dear Timothy,

Thank you for your message
  http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0008.html
on the OWL 2 Web Ontology Language last call drafts.

The WG did discuss the notion of ontology names and imports quite
extensively.   Some of the discussion is associated with WG Issue 15,
and is archived under
  http://www.w3.org/2007/OWL/tracker/issues/15

In the end, the WG decided that ontology names will be IRIs and that the
main way of accessing an ontology would be through this IRI.  (This
characterization ignores issues having to do with versions and series,
which is immaterial for the purposes of this response.)  

The use of IRIs as both name and location is in keeping with the
principles of the Semantic Web.  In the ideal case an ontology would
have a name, which must be an IRI, and would also be accessible at that
IRI.  It is the recommendation from the WG that ontology names and
ontology locations should be the same, and thus the WG is essentially
recommending that ontology import be by name.  

The WG does allow for ontologies whose names and locations diverge -
this is mostly for backwards compatibility - this situation will make
use and transmittal of ontologies more difficult.  As divergence of name
and location is not recommended by the WG *and* is against the
principles of the Semantic Web, the WG will not be providing special
guidance on how to deal with such ontologies.  Any such guidance would
require a separate name-to-location mapping mechanism.  This would be an
entirely new kind of mechanism for the Semantic Web.  It would give rise
to lots of problems, including uniqueness of names and authorities for
names.


The working group recognized that there are cases where an ontology
might not be universally accessible at the IRI which is its name or
where tools might want not to access the ontology at that IRI.  Even
though this goes against the principles of the Semantic Web, the WG has
published guidance for these cases, in 
  http://www.w3.org/2007/OWL/wiki/Syntax#Ontology_Documents
The WG left it up to OWL tools as to how they are to access ontologies
in ways other than IRI retrieval of the ontology name.  This is
appropriate, as different tools and environments may require different
mechanisms.

This solution is quite general and flexible.

1/ It allows for caching, i.e., where a copy of the ontology document is
   stored in some local document store, which is indexed by IRI.
   Caching can be done locally, in tools, or globally, e.g., in caching
   proxies.  Care should be taken to ensure that the cached documents
   are not stale, of course.  This kind of caching can be used for
   performance reasons and for off-line use.  It can also be used for
   ontologies whose documents are only accessible through non-Semantic Web
   mechanisms, e.g., through firewalls.

2/ It allows for local generation of ontologies without having to use a
   universally accessible name for the ontology.  Just use some IRI that
   is not (currently) globally accessible.  Ideally this IRI would
   become globally accessible when the ontology is ready for general
   use. 

The solution does not directly allow for transmittal of local names,
e.g., IRIs that retrieve different documents depending on where the
retrieval is performed from.  Such local names go against the principles
of the Semantic Web.  However, any use of local names, whether IRIs or
not, has the same problems.  Tools can, with some difficulty, overcome
locality by changing local ontology names both in ontology headers and
in imports.  This does require a knowledge of the source of the
ontology, but this would be required whether the names are IRIs or not.

The solution also does not directly allow for generation of ontologies
with local names that are later "moved" to other locations.  However,
tools can easily fix up ontology names and imports at the move time.
The WG does not feel that further guidance is needed in this area.


In general, the WG feels that its job is to support the use of OWL
ontologies that abide by the principles of the Semantic Web.  Any other
guidance is, at best, advisory, as there is no way to force tools to use
particular non-Semantic Web mechanisms.


The OWL WG does not plan to make any changes in response to your
comment.

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

Received on Thursday, 12 February 2009 15:24:39 UTC