Re: differences between OWL 1 and OWL 2

From: Sandro Hawke <sandro@w3.org>
Subject: Re: differences between OWL 1 and OWL 2 
Date: Wed, 01 Apr 2009 15:19:48 -0400

>> As far as I know, the changes from OWL 1 to OWL 2 are additions, with
>> only a very few exceptions.  The differences could be described as
>> follows:
> 
> Thanks.  A few clarifications, mostly by offering new wordings that I
> find clearer.  (I don't actually know if the claims are true; they just
> reword what I think you're saying.)
> 
>>   OWL 2 is almost entirely compatible with OWL 1, both syntactically and
>>   semantically.
>> 
>>   The functional syntax for OWL 2 is organized differently than the
>>   abstract syntax for OWL 1, but every construct in the OWL 1 abstract
>>   syntax has a directly corresponding construct in the OWL 2 functional
>>   syntax.
> 
> I wasn't thinking we'd include this kind of thing, although maybe it's
> helpful.  My concern is with OWL 1 ontologies which will be treated
> differently in an OWL 2 reasoner than they were in an OWL 1 reasoner.
> That's the "backward compatibility" issue, as I see it.

Fine.

>>   Just as in OWL 1, OWL 2 can handle all RDF graphs.  The vocabulary
>>   that is given special meaning in OWL 2 includes the special vocabulary
>>   of OWL 1.  However, the use of owl:DataRange is deprecated --
>>   rdfs:Datatype should be used instead.
> 
> For that last sentence, how about (if true):
> 
>     However, we now recommend the use of rdfs:Datatype instead of
>     owl:DataRange.  While owl:DataRange can still be used in OWL 2,
>     users should avoid it because ...(why?)...
      	    	   	    	    it is deprecated!

Deprecation is deprecation is deprecation.  :-)

> Are there other things you can still do that we're not suggesting not
> be done?  Maybe those should be listed separately -- they don't actually
> break backward compatibility.

This is the only "official" one.  There are lots of things that are
there for backwards compatibility and would probably not be there if
there had not been an OWL 1.

>>   The direct semantics for OWL 2 is almost completely compatible with
>>   the direct semantics for OWL 1.  The only difference is that
>>   annotations are semantics-free in the direct semantics for OWL 2.
> 
> add:
> 
>     This means that annotations no longer affect the entailments or
>     consistancy of an ontology; the fact that they did so in OWL 1 was
>     regarded by the OWL Working Group as an error.   

It is certainly the case that some people felt that annotations should
not have a meaning in either semantics, but I don't think that I would
go quite as far as your wording.

> Is there any (helpful) advice we can offer anyone who relied on this
> behavior of OWL 1?   

As far as I know this feature of OWL 1 wasn't implemented in many DL
systems for OWL 1, so depending on it would have been a bit silly.

>>   The RDF-based semantics for OWL 2 is completely compatible with the
>>   RDF-based semantics for OWL 1.  Some of the details of this semantics
>>   have changed, but the set of inferences are the same.
> 
> second sentence, replace with:
> 
>     The way the semantics has been specified has changed [to be
>     clearer?], but (except for the OWL 2 features), their meaning is the
      "to be less prone to paradoxes" is more correct
>     same.  This means the entailments of any RDF graph which does not
>     use the vocabulary introduced in OWL 2 are exactly the same
>     according to the OWL 1 RDF-based semantics and the OWL 2 RDF-based
>     semantics.
> 
> On the other hand, we might not need to say this, if it can be covered
> by a blanket statement (see below).

>>   The treatment of importing in RDF documents has changed slightly in
>>   OWL 2 if the RDF graphs are to be considered as OWL 2 DL ontologies.
>>   In OWL 1, importing happened first, so the entire merged graph was
>>   considered as one unit.  In OWL 2, the individual documents are
>>   considered separately in most cases.  This means that there are some
>>   groups of documents that could form an OWL 1 DL ontology but that do
>>   not form OWL 2 DL ontologies.
> 
> I'd want some more details or a link to more details here.  (And not
> just a link to http://www.w3.org/2007/OWL/wiki/Syntax#Imports but rather
> something that explains what I as user migrating to OWL 2 software need
> to worry about.)

The thing to say would be to talk about RDF documents that don't have
well-formed OWL 2 ontology headers in them, although headerless
documents are handled specially.

> So, if this is really all there is, then maybe we can have text like
> this in Overview:
> 
>     With the exception of two minor changes, OWL 2 is backward
>     compatible with OWL 1.
> 
>     {{EdNote||The Working Group considers these changes to be bug fixes
>     and expects them to solve more user problems than they might cause.
>     If they cause you any sigificant difficultly, please let us know.}}
> 
>     If OWL 1 users are not affected by these specific changes, they do
>     not need to change their ontologies when they start using OWL 2
>     software.  OWL 2 is defined such that all conformant OWL 2 systems
>     will behave just like conformant OWL 1 systems when handling OWL 1
>     ontologies.
> 
>     The changes between OWL 1 and OWL 2 which are not backward
>     compatible are:
> 
>        * In the Direct Semantics, annotations no longer have any
>          semantics.  They do not affect entailments or consistency. 
> 
>        * The treatment of importing in RDF documents has changed
>          slightly in OWL 2 if the RDF graphs are to be considered as OWL
>          2 DL ontologies. In OWL 1, importing happened first, so the
>          entire merged graph was considered as one unit.  In OWL 2, the
>          individual documents are considered separately in most cases.
>          This means that OWL 1 DL RDF documents that do not have a
>	   well-specified ontology header may need to be slightly
> 	   modified to be in OWL 2 DL.

Works for me.

>      -- Sandro

peter

Received on Wednesday, 1 April 2009 19:59:37 UTC