Re: Constraints explained (was: RIF needs different reasoning methods)

On Mar 13, 2006, at 1:39 AM, Michael Kifer wrote:
[snip]
> So, reasoning methods (e.g., derivation rules vs. constraints, query
> answering vs. query containment, etc.) are kosher annotations and 
> should be
> standardized to the extent possible (and reasonable) for
> RIFWG.

I fully agree with this. If you want to call these differences matters 
of "reasoning methods" rather than "semantics", that's also fine with 
me, though I find it a little jarring. (I remember Drew going on about 
this, but whatever.)

> Implementation strategies (e.g., top-down vs bottom-up) are
> non-kosher and should NOT be standardized.

Should be clear that I'm on board for this too :)

> That said, it doesn't mean that RIF should not have a mechanism for
> providing implementation hints (in an ad hoc way).

Or even a non-ad hoc way. Since I suggested this too, obviously I'm on 
board :)

[snip]
> So, if two parties come up with a way to communicate proof strategies 
> among
> themselves -- all the power to them. But we don't know what is THE 
> right
> way to represent and communicate such strategies (there isn't just one 
> way,
> clearly) so we shouldn't try to standardize this. We should just 
> ENABLE it.

Enable it in an extensible way, for sure. If there are existing hints 
in existing languages that we could use as test cases, so much the 
better. (As I said, I think both Jess and Jena have ways of indicating 
that certain rules should be processed top down rather than bottom up 
-- which if the rules aren't funky, I would hope would be semantics 
preserving (in some sense)...hmm...see:
	<http://www.jessrules.com/jess/docs/52/language.html#chaining>
Brrr. but I guess it's this sort of thing we must look too.)

By the by (and not particularly Michael directed), I know RuleML has 
all these rule types marked out...what's been their 
uptake/use/benefit/detriment?

Cheers,
Bijan.

Received on Monday, 13 March 2006 07:10:22 UTC