W3C home > Mailing lists > Public > public-swbp-wg@w3.org > January to March 2004

Re: [OPEN] and/or [PORT] : a practical question

From: Natasha Noy <noy@SMI.Stanford.EDU>
Date: Wed, 31 Mar 2004 10:42:32 -0800
Message-Id: <p0602041dbc90bd0f7fcd@[]>
To: Aldo Gangemi <a.gangemi@istc.cnr.it>
Cc: Ian Horrocks <horrocks@cs.man.ac.uk>, Guus Schreiber <schreiber@cs.vu.nl>, SWBPD list <public-swbp-wg@w3.org>, Nicola Guarino <guarino@loa-cnr.it>, claudio.masolo@ladseb.pd.cnr.it

Hi Aldo,

Indeed, I've read too much into your message. I apologize. The good 
news is we seem to be in almost complete agreement on this issue :))

>I do not want to recommend avoidance of metaclasses in general.
>I think there exist more than two positions in most group discussions.
>In this case, my position is a "qualified" agreement for best practices in the
>use of metaclasses (annotations, real metaproperties, etc.), together with an
>agreement on documenting worst practices in the use of metaclasses.

Well, we'll probably have to take it on a case-by-case basis (and it 
may be hard to impossible to agree on the worst practices. I'd rather 
focus on identifying the best ones). I am not sure what you mean by 
"qualified" but sometimes metaclasses are just a convenient way of 
modeling things and should be considered as one of the options (yes, 
with its trade-offs).

By the way, the discussion started with the issue that I wouldn't 
classify as a metaclass issue: how and whether to use classes as 
property values for instances (the original Bernard's question on 
dc:subject). To me, this is a separate issue (that later segued into 
the metaclass discussion), but it is also the issue of mixing classes 
and instances that takes us into OWL Full. In fact, lately, I find a 
lot more cases of having to go to OWL Full for this reason rather 
than to have  classes with other classes as instances (annotation 
properties do a good job of taking care of many of the cases in the 
latter category.)

>No, you are overinterpreting me. If avoiding metaclasses is less intuitive and
>cumbersome, it is not reasonable to put the burden on the shoulders of a na´ve
>I only suggest to describe *alternative* ways. These can then be used directly
>by the modeller if they have a comparable intuitiveness, otherwise can be used
>to generate mappings with appropriate tools, or simply suggested as
>alternatives (not better alternatives), in order to stay within OWL-DL.

we are in complete agreement here! :)

>Did I say that?!!! I intended that, from a scientific viewpoint, there is no
>stringent evidence that reasoning on metaclasses in the same problem space is
>*unavoidable* for the Semantic Web.

This is probably where we actually disagree (but I need to read your 
KR04 paper first). In any case, as you point out, this is a 
theoretical discussion and has nothing to do with best practices on 
the Semantic Web ( at least directly).

>New bottom-line: dear father and mother, I vouch for my humility.

but we knew that :)

Received on Wednesday, 31 March 2004 13:47:46 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 31 March 2004 13:48:53 EST