Re: Issues with classification a#b, a##b

Gary Hallmark wrote:

> Paul Vincent wrote:
> 
>> Gary - good point. But what would be the point of rules without the 
>> data/datamodel upon which the rules are based?
>>
> I don't know.  It seems kind of pointless to have a rule interchange 
> format without a data model interchange format.

? I was under the impression that in the business rules domain the data 
is typically exchanged via XML with associated XML Schemas. So even if 
you had a means to duplicate the data model in RIF you would still need 
the XML Schema as well in order to validate the instance data. Since you 
have to interchange the data model as a Schema anyway it seems kind of 
pointless to duplicate it.

> And because clearly the RIF needs to reference the data model, you can't 
> avoid tying the RIF to the "DMIF".  Therefore, you might as well talk 
> about at least a rudimentary data model in RIF.  

RIF needs to reference the data, so you need a means to access XML data 
from RIF, that's not the same as transmitting the data model itself via RIF.

> As I've been saying, 
> for my needs, some kind of frame type and generic lists are good enough.

>> Your scenario is more like "Oracle user tries to review a partial 
>> system transaction made up only of a RIF rules doc but missing 
>> associated schema reference" - I'm not convinced this is an important 
>> use case...
>>
> That's not a desirable use case.  That's what happens if RIF has no 
> usable data model.
> A "schema reference" is not a usable data model.  RIF rules reason on 
> uniterms and frames.  How do I turn a schema reference into statically 
> typed uniterms and frames?

Options:

(1) A standard mapping of XML Schema elements to frames

(2) An application-dependent mapping of XML Schema elements to frames 
plus a means to carry the Schema relationships themselves in RIF.

(3) A set of XML access and manipulation builtins

(4) A schema annotation mechanism so that the XML Schema (or some 
associated mapping document) can carry the information on how those 
schema elements map to the RIF predicates or frames. That's why I 
mentioned the SAWSDL spec which seems to give a minimal start on an XML 
vocabulary for doing that sort of thing.


For me a major issue is that you need your information to be strongly 
statically typed. A strong, static type system would be a major 
limitation on the applicability of RIF on the semantic web.

So if we can keep that strong, static typing out of RIF itself and in 
some associated XML Schema-specific layer then we have a chance of 
squaring that circle.

Dave
-- 
Hewlett-Packard Limited
Registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Tuesday, 31 July 2007 08:27:39 UTC