Re: draft response for LC comment 31

On 14 Feb 2009, at 11:07, Ivan Herman wrote:

> Peter,
>
> I am not 100% convinced by this answer, I think we should have some
> discussion either on the call or the f2f. Reading through Frank's
> comments, I do not find it unreasonable to have a dedicated annotation
> property that indicates the level of ontology which is intended in  
> spite
> of all the caveats that you describe.

Uh...we had long, extensive discussions of precisely this. I raised it  
as an issue thinking it would be a nice helpful thing. I came to  
believe that it's a very bad idea, at least given what we know.

> Of course, tools have to be
> careful not to believe this mark but it at least opens the door for  
> some
> reasonable conventions that the community could follow. The fact that
> some of the datasets/ontologies might be enormous is a compelling
> argument to have something like that around...

I don't believe that's remotely compelling. There's all sorts of  
properties about an ontology you may wish to mine or filter on:

	http://owl.cs.manchester.ac.uk/repository/browser

> Yes, this could be one of those extra small things that lead to hell,
> something you referred to at the last meeting. Nevertheless... I would
> like to have some discussions on this, if possible.

I object. This is not a "extra small thing that leads to hell". This  
is a feature that was discussed extensively by the working group and  
rejected:
	http://www.w3.org/2007/OWL/tracker/issues/111

I'm sorry, that Frank  finds the discussion in 111 "unconvincing" is  
not new information sufficient to reopen the issue. We were convinced.  
If Frank wants us to reconsider he must *argue* the case, not merely  
dismiss our discussion.

Remember I *raised* this; I *championed* this; I became convinced that  
any solution we came up with is unworkable. Prima facie we should  
respect that, and our own process.

(Note he doesn't raise the issue of imports at all :()

Furthermore, this is exactly the sort of thing that can be done, at  
lower cost and risk, outside the working group.

(The OpenCyc example is particularly silly...I don't believe it's  
particularly difficult to do profile checking on it, or on NCI, or on  
SNOMED...again, see the TONES repository.)

Cheers,
Bijan.

Received on Saturday, 14 February 2009 11:23:36 UTC