The cover material for the workshop shows the limitation of the current thinking in supporting resource discovery on the Web. While the workshop cover brings up and discusses important problems in the discovery of documents - a short term problem, it does not look at the problem for the long term - the web as the place to find and use any resource. In this position paper, we argue that the indexing schemes for the future web should consider two perspectives: Object Discovery and Meta Object Protocol. Our comments are based on our experience in the manufacturing sector and meta object protocols.
First, while the web has always been envisioned as an infrastructure where objects of various types (documents produced by humans being one of the types) can be found and used, the indexing schemes used today (e.g., www.lycos.com) are manifestly insufficient for objects which can be and have been formally described. If one considers examples such as manufacturing parts which can be categorized and described using Group Technology, engineering design documents which can be described using PDES/STEP standards, one realizes that many industries across the world use semantic indexing schemes that have been perfected for their use. Furthermore, services that are being standardized in the Object Management Group such as the trading service are an attempt by object vendors to support wide area discovery of program objects. Finding and using such objects is an important aspect of resource discovery in the WWW.
Second, one-size-fits-all approach to describing the semantics of meta information that can be exchanged between various repositories (such as the Dublin meta protocol) simply will not be used except for documents. Consider one of the fields in the Dublin meta protocol: language. If I write a Java applet what value should this language field contain - the Java language, Java VM code, or the native code generated by my Symantec compiler on my Intel machine or all of the above. Each of these uses of the keyword language are very important to a person trying to discover a Java applet. However, since the Dublin meta protocol considers human written documents, it does not allow for these nuances. In the long run, what will work is an approach the delivers a mechanism by which the meta information can be described consistently and the set of described entities extended gracefully. Putting the specifics of one set of semantics before thinking about a way to describe the semantics is a short term fix which will only proliferate the meta protocol standards achieving no convergence. What is needed is a meta object protocol - i.e., a formal way to describe objects about objects.
These observations are based on work that we are performing at
Crystaliz, Inc. using Knowledge Query Manipulation Language,
PDES/STEP part library standards,
and standards work proceeding at Object Management Group.
During the workshop we would like to present our experiences based
on our efforts in the formal objects part of the web.