RE: spec of multi-sorted Core should include the ASN productions

Hi Gary,

 

The method you describe to generate the Java meta-model of RIF is very
interesting and powerful. 

However, the way the XSD schema is generated from ASN07 is important in
that approach.

Thus, I think that the specification of this translation (ASN07-->XSD)
should be carefully designed in order to obtain a simple schema.

 

Analysing the structure you advise in the section 1.1 of 
http://www.w3.org/2005/rules/wg/wiki/Arch/XML_Syntax, such as the group
and the choice instructions, I wonder if those structures are widely
used in standard schemas, although they are closed and similar to the
DTD construction. 

 

Regards. Philippe.

 

________________________________

From: public-rif-wg-request@w3.org [mailto:public-rif-wg-request@w3.org]
On Behalf Of Gary Hallmark
Sent: Thursday, May 10, 2007 12:11 AM
To: W3C RIF WG
Subject: spec of multi-sorted Core should include the ASN productions

 

I was trying to build a little RIF translator for Oracle Business Rules
(OBR).  OBR is "strongly typed" like Java and so I want to use the
multi-sorted Core.  The "design" for my translator is

1. obtain an XML Schema document for multi-sorted RIF Core
2. use JAXB/2.0 to generate Java classes and xml marshaller/unmarshaller
from (1)
3. write a RIF "import" utility Java program that walks a RIF document
expressed using (2) and creates production rules in OBR
4. write a RIF "export" utility Java program that creates a RIF document
from an OBR ruleset (or explains why translation isn't possible)
5. add some builtins from xquery/xpath like op:numeric-add which are
sorted and without which it's kind of hard to write any useful rules at
all

It is reasonably straightforward to generate (1) from the ASN in the
Core spec, although the Schema isn't uniquely determinable from the ASN
(possibility of stripe skipping, etc), which is a big problem for
interoperability.

Unfortunately, the ASN in the Core spec does not include support for
multiple sorts.  The "sorting" is kind of "added on" toward the end of
the spec, and the ASN is never completely revealed.  I think it's pretty
easy to guess about the primitive sorts from the examples, but the arrow
sorts and boolean sorts really need to have ASN.  

-- 


  <http://www.oracle.com> 
Gary Hallmark | Architect | +1.503.525.8043
Oracle Server Technologies
1211 SW 5th Avenue, Suite 800
Portland, OR 97204