W3C home > Mailing lists > Public > public-owl-dev@w3.org > July to September 2007

Re: Pellet gives strange, wrong error with Booleans

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Thu, 30 Aug 2007 19:51:06 +0100
Message-Id: <C3D18423-248D-4784-AC26-F658DB452FA9@cs.man.ac.uk>
Cc: Owl Dev <public-owl-dev@w3.org>
To: "Stephen Larson" <slarson@ucsd.edu>

Cross posting trimmed. By saying I wanted you to move it to an  
appropriate public forum I did not thereby suggest that you go on a  
cross posting spree. A bit of sensible netiquette really goes a long  
way. As does, of course, simple etiquette.

On 30 Aug 2007, at 19:19, Stephen Larson wrote:

> Bijan,
>
>    Sorry for not using the appropriate channels before.  I have  
> included other lists on this because I think that it is important  
> to clear this up.
>
>    I'm sending you the simplest example of this problem that I can  
> find.  It doesn't seem to happen with a single simple case of the  
> presence of a boolean.  The behavior appears to be more complex  
> than that.  But it is the case that the problem exists, I've sent  
> you an example ontology that exhibits it, and it can be reversed by  
> commenting out the booleans.  It is reproducible if you use the  
> tools that I have described.
>
>    I understand if you are frustrated with the DIG interface.

I don't think you do.

> But, it seems like you are implying that Pellet doesn't really  
> support the interface.

Well, no. But I say now that it seems that you don't really  
understand the limitations of the DIG (1.1) interface.

	<http://dl-web.man.ac.uk/dig/2003/02/interface.pdf>

"""This Level 0 Version 1.1 provides rather restricted support for  
concrete domains.
Integers and strings are provided, along with concept expressions for  
minimum, max-
imum, value equality and ranges. Linear inequations are not provided,  
nor are
named concrete objects, although these will be introduced in a later  
version of the
schema. Ranges can be asserted for attributes using assertions like  
rangeint or
rangestring. If no range is supplied, attributes have integers as  
their range by
default.
"""

The error you got, IIRC, claimed that that "False" is not an "int",  
which implies (to me) that the interface was coercing an unknown  
datatype to a known one. I admit that I haven't verified this  
directly, but you aren't encouraging me to do that work.

As far as I know, Pellet is DIG compliant. It's possible that there  
are bugs, but I believe that what you are hitting is a known problem  
in the DIG specification.

>    I'm just a naive user here with limited resources.

Whereas I have unlimited resources which I'm more than glad to put at  
your disposal.

>   I'd like to use Pellet to check out my ontology in Protege 3.3.1  
> without having to acquire a 'toolchain' in order to do it.

Protege3.3.1 and Pellet *are a toolchain*. DIG is known to have issues.

Dude, just test it out with FaCT++ under dig and see what you get. My  
bet is a similar response.

> The pellet website says it supports DIG: http://pellet.owldl.com/ 
> faq/protege/.  But it seems like maybe it doesn't really.

Eh, your insinuations grow wearisome. I pointed you to more robust  
alternatives. I could point out that DIG 2.0 is underdevelopment to  
address these and similar issues.

>   So it seems like the DIG interface is really just a distraction  
> and shouldn't be included in code intended for naive end-users.

I never recommended to you that you use Protege 3.3.1 and DIG with  
*any* reasoner.

>   Maybe some agreement between the Pellet and the Protege people to  
> stop supporting the DIG interface would be the best route to avoid  
> naive users like me from falling into these kinds of traps, getting  
> misleading error statements and thus being unable to fix their  
> ontologies so that Pellet can reason with them.

Take it up with the Protege 3.3.1 folks. Protege4alpha shows that you  
can have a more reliable connection with a reasoner than DIG 1.1. I  
showed you several ways to work with your ontology for reasoning,  
e.g., OWLSight, which is better in a lot of ways as it supplies  
explanations as well.

>   I think that would be best for the community.

Yes, I'm sure you do think that.

(If you wanted to do something constructive, you could add  
documentation to the Protege Wiki about limitations of the DIG  
interface...or even just start it, and solicit people to help. That's  
MUCH more constructive than this obnoxious, holier-than-thou, my- 
needs-are-superimportant moaning.)

Cheers,
Bijan "And of course I'm fallible and could have made a mistake in  
what DIG says or what's happening" Parsia.
Received on Thursday, 30 August 2007 18:49:56 GMT

This archive was generated by hypermail 2.3.1 : Wednesday, 27 March 2013 09:32:55 GMT