W3C home > Mailing lists > Public > public-owl-wg@w3.org > January 2008

Re: Primer review, part 1 Introduction, Orientation sections

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Mon, 21 Jan 2008 12:45:32 +0000
Message-Id: <2E3C5313-549F-4764-82E8-498DD8F57CDA@cs.man.ac.uk>
Cc: OWL Working Group WG <public-owl-wg@w3.org>
To: Alan Ruttenberg <alanruttenberg@gmail.com>

Thanks for the feedback on our first draft. I take it that you do  
like the technology perspective approach, at least in principle.
On 21 Jan 2008, at 06:47, Alan Ruttenberg wrote:

> These comments are on the Primer document http://webont.org/owl/ 
> documents/primer.html as accessed late night 1/20/2007. They are  
> detailed comments on the Introduction an Orientation sections.  
> Review of other sections will follow.
>
> Summary:  There are a number of places where I think the  
> descriptions don't adequately introduce OWL from the point of view  
> of someone embedded in the technology described, which I think is  
> important for drawing readers in. Ideally the orientations would  
> make affirmative statements about why people coming from  
> experiences with their technology might be interested in OWL, and  
> what they should not expect to find.

We need to take care to balance breath and depth. I believe we can  
expect to "set the frame" adequately with this document, but then we  
should hope that people are inspired by this section to write more  
detailed perspectival discussions, or to go to lists, etc.

So some sense of how much more text you think is reasonable for these  
intro sections would be good.

Also, I think scattering some "NOTE FOR RDFERS: You can see blah balh  
balh" through the example discussion could be more helpful than  
adding a lot more upfront discussion.
[snip]

> Introduction doesn't mention important bits about design  
> considerations for OWL - that it there are known algorithms that  
> can give complete answers etc, and that is is designed to the most  
> expressive language for which one can do this, and why it is  
> important to aim for this.

I'm not sure foregrounding this in this way is all that helpful. I'd  
rather talk about "yes and no" answers along the way. I mean,  
variable free syntax is an important design consideration as well  
historically, but I don't see it's good to *foreground* that.

[snip]
> "both OWL and XML have an object oriented approach"
>
> I don't consider OWL to be object oriented,

This would seem to be a very much minority view.

> and think this will confuse rather than help.

Well, the point is for both XML and OOP sections are to say, roughly,  
Yes we have classes and objects, but don't rely on your understanding  
from other class/object pardigms. Actually, I think it's less  
important to say *what* the differences are than to make clear that  
there *are* radical differences. I.e., to "set the frame". If we say,  
"Whoa they are radically different" and the person doesn't get how,  
they are in a much better position to ask questions than if we  
*don't* say that they are radically different and they don't think to  
ask certain questions.

In other words, I'd prefer that the fallback mode, in general, be  
"self aware" confusion than "unknowing misunderstanding".

[snip]
> Database use does not imply negation as failure. It doesn't offer  
> any support one way or another it seems to me,

Again, this is not a particularly standard view. Canonically,  
relational dbs make the CWA and NAF. Consider how aggregation works.

> whereas OWL offers support for querying over information that is  
> not explicit and not fully specified.
>
> There is no mention of very important difference, namely that there  
> are no integrity constraints in OWL.

It's not clear to me that we should call this out as such, esp. up  
front. I'd rather have a discussion at the QCRs.

[snip]
> I don't think this will reach typical database users. It is too  
> theoretical. It needs to be more down to earth. Clearly there are  
> very obvious differences - transactions, triggers, replications - a  
> whole set of considerations that OWL does not deal with.

Some of these are features of *systems* not the modeling formalism.  
It'll probably help to make that distinction a bit, though again, I'd  
prefer to be a bit subtler. Discussing ER diagrams might be more useful.

[snip]
> Object-oriented Programming.
>
> Again, focus on information completeness won't resonate with oo  
> types. First they have to be told the "further stuff" and then why,  
> despite this, they might be interested in OWL.

I intend to stress the "Template" vs. "description" perspectives  
(plus the lack of behavior). This will need to be synched more with  
the XML and RDBMS discussions. The RDF discussion perhaps should come  
last.

In my discussions from people coming from XML/OOPy background, this  
has been the most helpful for them.

Cheers,
Bijan.
Received on Monday, 21 January 2008 12:43:38 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 21 January 2008 12:43:40 GMT