RE: http://www.w3.org/2006/11/04-rif-minutes.html#action04: Notes from UCR breakout of 11/04 --> coverage

Hi Paul,
 
Just a quick response to points 1 & 2.
 
When I said it "must be possible" to implement a rule engine for a RIF
dialect, I did not mean this as a "requirement," rather as something
that follows from the fact (requirement) that every RIF dialect must
have precise syntax and semantics.  As you say, it is a "side effect."
Since the whole idea is to give an "intuitive" and "informal"
definition of "coverage" at this stage, I think it is ok to talk as
though RIF is executable by some appropriate piece of software.
 
Allen


________________________________

	From: Paul Vincent [mailto:pvincent@tibco.com] 
	Sent: Wednesday, November 22, 2006 12:46 PM
	To: Ginsberg, Allen
	Cc: public-rif-wg@w3.org
	Subject: RE:
http://www.w3.org/2006/11/04-rif-minutes.html#action04: Notes from UCR
breakout of 11/04 --> coverage
	
	

	Thanks Allen:

	 

	Yes we probably need to agree to disagree and move to
arbitration! So I cc'd the rif-wg for other comments on this topic.

	 

	My thoughts on your comments below are: 

	1.	As RIF is an interchange format, the fact that a RIF
dialect may be considered a rule language must be a side effect - and
therefore should not be reflected in the definitions. In other words it
is not necessary for RIF dialects to be a rule language. 
	2.	Why must it be possible to implement a rule engine for
a RIF (format) dialect? For example, I can use XML as an interchange
language for representing SQL queries. However, this does not mean I
*need* to have database engines that can execute said XML. To me, this
requirement seems somewhat orthogonal to the charter. 
	3.	Mentioning software is also orthogonal. For example,
why cannot a rule translation be handled manually, like it may be done
today and will be done for initial examples I am sure? Again, this does
not seem to be a requirement (although of course it *is* a requirement
that the process is automatable). 
	4.	Circularity: possibly my definition was circular.
Perhaps Coverage should imply translation to, from or to+from a RIF
dialect. 

	 

	An updated version of my coverage proposal would therefore be:

	*          "RIF covers a rule language" if
	1. rules written in the covered rule language can be translated
to, from, or both to and from a dialect of RIF 
	and 
	2. rules mapped from 1 language to another via a RIF dialect
exhibit equivalent behavior

	 

	Note: I did not use "identical" behaviour as it is not the
remit of RIF to standardize rule engine outputs. So if I have 2 rules:

	If customer is new then customer credit period is 1 month

	If customer is new then customer credit period is 1 week

	Then the equivalent behaviour is to either assign a customer
credit period of 1 month or 1 week (depending on the rule engine).

	 

	Cheers

	 

	Paul Vincent

	TIBCO - ETG/Business Rules 

	 

	PS: my original counterproposal sent as PPT was:

	*          "RIF covers a rule language" 
	means 
	1. rules written in the covered rule language can be translated
into 
	a dialect of RIF 
	and 
	2. the resulting rules represented in RIF can then be mapped to
some target rule language, other than the original language, to achieve
the same behavior as the original rules

	 

	
________________________________


	From: Ginsberg, Allen [mailto:AGINSBERG@imc.mitre.org] 
	Sent: 22 November 2006 17:15
	To: Paul Vincent
	Subject: RE:
http://www.w3.org/2006/11/04-rif-minutes.html#action04: Notes from UCR
breakout of 11/04

	 

	Hi Paul,

	 

	I see your point, but I don't agree.   For many intents and
purposes a RIF dialect should be considered a rule language.  Perhaps
no one will ever implement a rule engine for a given a dialect, but it
must be possible to do that.  Moreover,  if we don't talk about
"software that works with RIF" then we need to say that "coverage"
means translating from L to other rule languages in the same dialect
that are also "covered" by RIF.  But that is clearly a circular
definition, i.e.,

	 

	           when we say that "RIF covers rule language L" we
mean that any rules written in L can be translated into a dialect of
RIF and the resulting RIF document can be translated into any rule
language covered by

	            the same RIF dialect...

	 

	Allen

	 

	 

		
		 

		
________________________________


		From: Paul Vincent [mailto:pvincent@tibco.com] 
		Sent: Tuesday, November 21, 2006 3:47 PM
		To: Ginsberg, Allen
		Subject:
http://www.w3.org/2006/11/04-rif-minutes.html#action04: Notes from UCR
breakout of 11/04

		Hi Allen - 

		 

		Re ACTION: PaulV to work with Allen on defn of covers
[recorded in http://www.w3.org/2006/11/04-rif-minutes.html#action04]

		 

		I was trying to recall the discussion on Coverage from
the F2F and all I had was (a) your note below and (b) my PPT edits
(attached).

		 

		I note my edits attempted to add (some of) the vigor
you were trying to avoid... oh well.

		 

		Looking at your text again:

		Intuitively, when we say that "RIF covers rule language

		 L" we mean that rules written in L can be translated
into a dialect of

		RIF 

		Which is OK

		and the

		 resulting RIF rules can be used by software designed
to work with RIF

		This seems redundant - why refer to external software
in a definition of coverage?

		 to achieve essentially the same functionality as
enabled by the

		 original L rules.

		This part is fundamentally incorrect for a rule
interchange format. The behavior described is a corollary of a rule
being interchanged from 1 language via RIF into another language.
Indeed I dislike the term "RIF rules" as its open to misinterpretation
implying RIF is a rules language, and is ambiguous with transformation
rules (to and from RIF).

		 

		Any thoughts?

		 

		Paul Vincent

		TIBCO - ETG/Business Rules 

		 

		 

		 

		-----Original Message-----
		From: public-rif-wg-request@w3.org
[mailto:public-rif-wg-request@w3.org] On Behalf Of Ginsberg, Allen
		Sent: 05 November 2006 13:18
		To: public-rif-wg@w3.org
		Subject: Notes from UCR breakout of 11/04

		 

		 

		 

		[PV>] ...

		 

		PROPOSED TEXT ON COVERAGE:

		 

		 

		 We note that in this document we deliberately refrain
from defining

		the

		 notion of "coverage" in a rigorous manner, since
precisely what it

		 means for diverse rule languages to be "covered" by
RIF may vary from

		 case to case.  Intuitively, when we say that "RIF
covers rule language

		 L" we mean that rules written in L can be translated
into a dialect of

		RIF and the

		 resulting RIF rules can be used by software designed
to work with RIF

		 to achieve essentially the same functionality as
enabled by the

		 original L rules. 

		 

		[PV>] ...