- From: Michael Kifer <kifer@cs.sunysb.edu>
- Date: Thu, 23 Jun 2005 00:54:52 -0400
- To: Jim Hendler <hendler@cs.umd.edu>
- Cc: Bijan Parsia <bparsia@isr.umd.edu>, www-rdf-rules@w3.org, dreer@fh-furtwangen.de, Jos de Bruijn <jos.debruijn@deri.org>
Jim, So, you are saying that LP is at the same stage as DAML+OIL before standardization. What is needed is to work out the details -- typically a job for a working group. Who disagrees with that? And you didn't need to defend OWL because nobody was attacking it. Our discussion was about the 1-stack vs. multi-stack architectures. Since in a previous email you said that you are "not against a multi-stack solution" then I don't see what is the point of contention (at least at the high level). I believe that everybody agrees that the more integration -- the better. We just don't believe that integration of the leading useful technologies to the point that only one stack is left is possible. --michael > Mike - I think you're missing the point of the "webbie" nature of OWL > and the difference from traditional KR, but I have written that up > too many times to do it again here. With LP, the question is how I > can use your rules/program/etc. in part to get a "network effect" and > to make it so I can link together the logics and logic programs as > easily as I link web pages. It;s not that no one has good research > ideas on how to do that, it's how to bring those to fruition and > greater use that is the key > The ontology stuff in OWL, which is not actually DL (even OWL DL > departs from traditional DL in some interesting ways, but OWL Full is > the one I care most about) is based on many years of work in AI, and > was explored on the web long before OWL was done - cf the SHOE work > my group did (still a high hit at Google - so just google "shoe") and > that was followed by XOL, OIL, and others before the standardization > began. > I think the LP stuff is in similar state - a basic idea has been > fleshed out, some variants are being explored, and there is a govt > interest in pushing for a de facto standard. But going from there > to the finish line is where a lot of the time and blood goes in -- > it's in making the stuff fit with what else is out there in the Web. > We had to do a lot of work to make OWL fit in with RDF and other > languages it needed to interoperate with, and a web rules language > needs to be defined with the other things already in the space it > wants to play in (thus the "stacks" issue - if it wants to be in a > Sem Web stack, it needs to play with other SW stuff; if it wants to > be in the XML stack, it needs to play nice with XML stuff like Xquery > and Xpath, etc. > And that is the discussion we are having -- but if we can nail this > stuff, the result is worth it -- OWL is certainly the most used > KR/ontology langauge in the history of AI as best anyone can tell, > and if we want the Web Rules Language to flourish we want it to grow > like the Web does, not like rules languages have -- nothing wrong > with the latter, but there's a whole lot more Web pages out there > than logic programs, and it's a lot more fun to play in the > exponential growth space :-) > anyway, we're all working for same ends, just different means, and > finding the consensus space in the middle is wondefully non-fun, but > worth it in the end > > Ok, end of crap, back to technical issues and Greek symbols... > JH > p.s. please note - I spent many years of my career arguing against DL > and doing scruffy AI - yet here I am defending OWL - why? because > the design time and fights over the details of a number of use cases > ended up creating something pretty damn useful -- both in the OWL DL > space and in the OWL Full space -- so somehow the process worked... > > > At 20:41 -0400 6/22/05, Michael Kifer wrote: > >Jim Hendler wrote: > >> > >> Mike - I think you misunderstand the stuff about stacks and etc -- I > >> hope my use cases (in public-sws-ig@w3.org for those just joining the > >> conversation) would help make it clear that these are not separate > >> and unrelated stacks, nor are they identical things -- the key is > >> figuring out how the stacking works and how things interact -- I'm > >> not against a "multi-stack: solution, but as far as I am concerned > >> the more overlap the better, and I am fairly sure that we can do > >> significantly better than DLP in terms of providing a useful web > >> rules language that interacts well with the existing, and becoming > >> more widely used, ontology spec.* > > > >I think it is not just me, but a number of people who read your paper on > >the two stacks may have misunderstood it. At least one way to understand > >what is said there is that 1 stack is good and 2 is not. > >If there is another way to understand it (as advocating a multi-stack > >architecture) then this second meaning is deeper than I was able to dig up. > > > >> Seems to me the key is exploring how to get maximum interoperability > >> between the important work in BOTH areas (and I defy you to go back > >> through this discussion and find any email where I haven't said I'm > >> in favor of a web rules language) > > > >You didn't say this and I didn't say that you said this. I was > >focusing on what I think were technically inaccurate claims in your email > >regarding the layering of WRL on top of DLP (where WRL is taken to mean the > >particular language under this name and not "a" generic web rules language). > > > >> and also how to get the Web rules > >> to join in the growing whole that is the semantic web -- it's not the > >> same as applying LP in the Web area -- I argued for nearly a decade > >> about the difference between Web Ontology and standard AI KR > >> languages, and OWL has some significant differences from traditional > >> AI (see the OWL FAQ [1] and the discussion of KR back in the 2001 > >> Scientific American article [2]) > > > >Not "applying LP in the Web area" but "adapting LP to the Web". > >Technically, OWL is an adaptation of DL to the Web with some additional > >research needed to accommodate RDFS. But in the LP area this research has > >already been done years ago. > > > >> This latter, btw, explains why URIs > >> are not just some wildassed thing, they're crucial to the Semantic > >> Web in a very deep way - read the Sci Am or any of Tim's discussions > >> of this issue. > > > >Of course URIs are crucial. After all, they are object identifiers, so they > >are as crucial as any notion of an Id. > > > >But do they imply/require a new kind of KR? There are interesting new > >problems that stem from the architecture, but don't make it sound as if the > >"old KR" is out of the window and adapting it to the new architecture is a > >hard or pointless exercise. The LP paradigm is as applicable to the Web as > >DL, if not more. (I, of course, think that it is more :-) > > > > > >> so, I don't see this as in any way being a discussion of rules vs. > >> ontology -- in fact, I cannot think of any dumber way to approach it > >> -- rather it seems to me we're trying to explore where these things > >> can overlap to the benefit of users and of the Web -- that strikes me > >> as a very worthwhile pursuit > > > >The term "rules" is ambiguous in the context of our discussion. If you said > >"I don't see this as in any way being a discussion of *LP* vs. ontology" > >then this is exactly what I was trying to say. As I remarked above, > >the 2tower paper **appears** to be arguing that LP+OWL in a 2-stack > >architecture is a nonstarter. > > > > > > --michael > > > > > >> -JH > >> > >> > >> [1] http://www.w3.org/2003/08/owlfaq.html > >> [2] > >> > >>http://www.sciam.com/article.cfm?articleID=00048144-10D2-1C70-84A9809EC588EF21 > > -- > Professor James Hendler Director > Joint Institute for Knowledge Discovery 301-405-2696 > UMIACS, Univ of Maryland 301-314-9734 (Fax) > College Park, MD 20742 http://www.cs.umd.edu/users/hendler >
Received on Thursday, 23 June 2005 04:56:33 UTC