- From: Phil Tetlow <philip.tetlow@uk.ibm.com>
- Date: Thu, 14 Oct 2004 08:22:18 +0100
- To: "Jeff Pollock" <Jeff.Pollock@networkinference.com>
- Cc: "SWBPD" <public-swbp-wg@w3.org>, <eric@w3.org>, <craig.gannon@carbacle.com>
Jeff Many thanks for such an inspired response to my somewhat clumsy thoughts. I think that you have managed to encapsulate to number of very important issues here that require genuine open debate within a number of communities in our industry – not just the W3C working groups. Having the conviction to willingly collate and communicate such opinions indeed deserves recognition and credit. I agree with most of your points and, in particular, would like to reinforce:- - Right or wrong, good or bad, there still needs to be a common base from which real-world user communities can start to investigate and utilise the benefits now afforded by the Semantic Web. Hence, I also think that the DAWG charter has hit the right spot. Maybe XQuery is not the ideal straw man, but it is definitely one of the strongest competitors in the race. Building a pragmatic base from foundations that have already been proven and experienced mass market take-up appears absolute common sense to me. This should not, however, constrain work moving forward in new fields of technical endeavour that are more closely aligned with purist interpretations of underlying technologies. - Does this ‘base’ necessarily need the blessing of organisations like the W3C? I guess only time will tell. Nevertheless, naivety aside, I consider that supporting the principle and spirit behind a number of ‘non-gold-standard’ Semantic Web technologies should be a very valid agenda item, so long as those vendors involved are prepared to follow appropriately open protocols and decorum. - Perhaps I am still somewhat unique in having experienced real world demand for ‘industrial strength’ Semantic Web technologies, but the need is still very real and, from where I’m currently standing, appears to be picking up momentum fast. For this reason there seems an urgent need for those who can make a difference in our profession to send out some strong collective guidance to a range of user communities. Is this a job for organisations like the W3C alone? I’m personally not sure, but I agree that lack of significant vendor participation – for whatever reasons – in appropriate standards initiatives may well work against us all in the long run. I am resigned, for now, to concede that we are were we are! Finally I absolutely sign up to your grey view of the world and also believe that when industry, academics, and standards bodies work synergistically - great things can happen. I also hope that our exchanges over the past couple of days will encourage other W3C members to comment. I feel sure that there are still some interesting opinions that have not yet been voiced, but which could matter a great deal. Kind regards Phil Tetlow Senior Consultant IBM Business Consulting Services Mobile. (+44) 7740 923328 "Jeff Pollock" <Jeff.Pollock@net workinference.com To > Phil Tetlow/UK/IBM@IBMGB cc 13/10/2004 19:14 "SWBPD" <public-swbp-wg@w3.org>, <eric@w3.org>, <craig.gannon@carbacle.com> Subject RE: NI contact re: XQuery & DAWG Phil- Thank you for your thoughtful reply and your willingness to share the topic of discussion more broadly. Like you, we at NI view the standardization efforts for a query language specification as being of the utmost importance to the broader Semantic Web initiatives. In my honest opinion, there is more than a passing resemblance to ANSI SQL efforts in the early 90's. Like SQL92 did for relational systems, someday a query specification will unify fractured implementations and ease the adoption risk for commercial buyers of the Semantic Web stack. SPARQL/DAWG, on its current path, will not unify the Semantic Web stack, nor provide the launchpad for widespread adoption of the broader vision. Arguably, the DAWG never had had such a grand charter - to unify and align fractured SW specifications with opaque relationships to one another. DAWG is almost entirely concerned with an RDF triples "database interrogation patterns." However, the DAWG charter does/did have *some* alignment and unification goals in mind: (from DAWG charter) "There is a requirement for RDF data to be accessible within an XML Query context. [...]While the data model of the query language of this protocol is dissimilar to that of XQuery, a non-XML concrete syntax might reuse syntactic elements from XQuery to aid learning time, even if XQuery is not chosen as the strawman." Regardless of nits about the DAWG initiative, and the resulting SPARQL, which I will spare you a litany of critiques on - I'd rather focus on the positive points for why NI's position on XQuery is indeed better for the industry as a whole. First, from a "what's good for the end user" perspective: * those who want to adopt, but lack confidence in supportability * those who have tools with XQuery interfaces today * those who have XML centric initiatives they want to build upon * those who want to adopt SW, but don't want to retrain * those who have fought for, and won, the XML migration battle Second, from a "what's good for the W3C" perspective: * create a semblance of consistency & unification for the SW stack * adhere to layered guidelines (grammar separate from algebras) * create a common façade to entire SW stack * improve adoptability by wider community * recognize vendor & customer investments in XQuery to date Third, from a "what's good for the systems integrator" perspective: * SI's using XQuery based middleware tools * SI's using XQuery based XML repository technology * SI's who have trained thousands of developers on XQuery * SI's who have pushed for XML-centric SOA customer solutions * SI's who are already coping with too many one-off languages Fourth, from a "what's good for the vendor community" perspective: * ISV's who want to protect their past investments in XQuery * ISV's who don't want the market further confused by query "noise" * ISV's who want to implement consistent interfaces to data layers * ISV's who want to use OWL alongside RDF and XML Now, there are nuanced arguments that can be (and have been) made about XQuery alignment with RDF data structures (triples, not taxonomies), and there are concessions that will be trumpeted as a form of support (XML result sets), and there can be debates about community support levels (SQL not XQuery) -- but these all miss the point. The point is that there are a variety of data formats in the SW stack (XML, RDF, OWL, and rules all have fundamentally different theoretic formalisms) and that there a wide variety of different usage patterns for SW technologies (databases, application servers, middleware, repositories, registries, content aggregators, search tools, etc.) and that there is obviously no one single query algebra/model that can meet the needs of such a diverse array of formalisms and usage patterns. BUT, we CAN agree to use a common GRAMMAR! As it turns out, the built-in mechanisms for extensibility of the XQuery grammar are a near-perfect fit for overlaying variant underlying algebras - NI has shown how this could work in our XQuery grammar proposal to DAWG - not to mention all other good reasons identified in points 1-4 above. Phil, you appropriately point out the deep divide here between the "purist view" of the world vs. the "pragmatists view" of the world. Personally, I am most interested in how to make the snowball (of adoption) roll faster -- I believe that SW alignment with XQuery grammars is one of the best things that we as a community can do to advance that cause. You also point out the distinction between what standards efforts are good for vs. what industry alliances good for. I don't think the distinction is as black and white for me. In principle, I believe that when industry, academics, and standards bodies work synergistically - that great things can happen. Unfortunately, this is not happening within DAWG. For instance, why aren't Oracle, Microsoft, Software AG, BEA, IBM, or others working in this group? It seems to me that the software marketplace has invested hundreds of millions, if not billions, on the W3C XQuery initiatives - all of the vendors mentioned above are shipping products and training people on XQuery today. I would think that they should have an interest in a new - non XQuery compatible - query standard coming from within the W3C. I would also think that DAWG and the W3C would be more interested in gaining consensus from vendors who have invested fortunes in query standard efforts for XML, which is in fact the bottom layer of the SW stack. Network Inference supports an XQuery grammar interface to OWL, and more broadly, for the rest of the stack including XML, RDF and a future rules specification. This is our position, it is above-board, and it is based on solid technical, business, and community interests and rationale. Thanks again for the opportunity to restate our thoughts on the matter, I look forward to an ongoing discussion - either in public or in private. Warm Regards, -Jeff- -----Original Message----- From: Phil Tetlow [mailto:philip.tetlow@uk.ibm.com] Sent: Wednesday, October 13, 2004 2:38 AM To: Jeff Pollock Cc: SWBPD; eric@w3.org; craig.gannon@carbacle.com Subject: Re: NI contact re: XQuery & DAWG Jeff Pollock (DAWG) wrote: I'd like to gauge your interest, motivation, and timing for the XQuery to RDF/OWL initiatives Jeff, I'm pleased to discuss, but, unfortunately, can only offer my personal opinion. You will be aware that I have some limited contact with the DAWG and am always interested in the ongoing debate in this working group, having a long history with traditional relational database technologies. As such I consider semantic web query language standardisation to be an important issue. I think that there are really two separate issues here:- · What should be recommended as a best practice approach for Semantic Web data interrogation - I see this more as a standards issue and something that should not, necessarily, be of direct concern to vendors · What implementation methods should be advocated for real-world early adopters - This, obviously, need not be of great relevance to appropriate standards bodies, but should be of great importance to vendors in order to establish market growth On the first point you may be aware that there has been some recent discussion in the Best Practices Working Group which advocated approaches based around the embodiment of the data sources being interrogated - The Semantic Web has, to date, been formed on data structures derived from a triple based representation. As you know, there is good reasoning behind this and hence so, to follow the argument forward, it makes absolute sense to me to use a querying mechanism formed from the same structural roots. XQuery , unfortunately, does not naturally meet this criteria and I hence align with the triple-based query school of thought. Sorry, this simple boils down to the old adage, 'the right tools for the right job'. Nevertheless, you would never hand a surgeon a 3 inch scalpel on his first day at medical school! So, I also believe that there is also some room here for a practitioners view. Having worked for a year and a half on a significant real-world project with a desperate need for a large underlying formal ontology, I have experienced the Semantic Web's growing pains first hand and fully appreciate that your average 'technician' finds a purist view of the Semantic Web too abstract at present. Personally I think this is a problem with the educational system lagging too far behind leading edge concepts - Most of the fresh Computer Science graduates I encounter are still steeped in traditional relational data implementation, with very few even being close to understanding the theory behind good relational design. Ask them what a tuple is and they immediately talk about flowers from Holland!! Trying to discuss the value of minimised data representation via triples is pointless. For no other reason than Semantic Web technology take-up, I hence feel that there is also a need for vendors to align and decide on 'interim' query solutions that have a much closer syntax to good old SQL. Again I do not consider this to be an issue for the standards bodies. Will this 'quick and dirty' approach to semantic querying have a detrimental impact on the Semantic Web? - I think not. There will always be other, purer solutions based around triples and I think it would be wrong for vendors to fight against these. Overtime I am sure that they will be incorporated into vendor specific implementations and toolsets, much the same way that object-oriented database technologies sneaked into mainstream database products. But at least at that point the consumer will be given the option to choose. Addressing product specifics via standards lobbying appears, to me, to be the wrong tack. Forming industrial alliances to promote market growth is another matter and I am more than happy to act as a conduit within IBM for you to open up discussions with the most appropriate in our Software Group for this process to be discussed. I would, however, appreciate your cooperation to ensure that appropriate levels of commercial decorum are maintained. The Semantic Web is an important contribution to global technology progress. Nevertheless it is still a child that we all passionately want to care for, and unfortunately, children have a propensity to learn bad habits from their guardians - a failure of which I am acutely aware. I'm not so bothered about guardians being misled as they should be worldly-wise enough to make up their own minds. Kind regards Phil Tetlow Senior Consultant IBM Business Consulting Services Mobile. (+44) 7740 923328
Received on Thursday, 14 October 2004 07:19:52 UTC