On Monday 18 February 2002 12:14, Kimbro Staken wrote: >. . . > After all we're talking about an XML database here, > metadata is supposed to be inherent in the format. >. . .
Yes, but if you're considering implementing versioning in the database (I don't know if it's the case), it would be a big advantage in many cases to be able to find out which "fork" (data, metadata, etc.) of an element was modified. For example, knowing that content wasn't modified but simply promoted to "status=approved" can be very important. Actually I didn't think of that before, but versioning info/history is definitely another type of "fork" on the data. These last two points make me think that the decision to implement user metadata as part of the database is strongly related to the decision about versioning. > Hmm, that actually brings up an interesting point for system level > metadata too. Seems it should be queriable via XPath like the rest of > the document content. I guess you can present the data as virtual > namespaced attributes on the document. Maybe with an option on > retrieval to get the meta data physically added to the document. Sounds interesting, goes along with the idea of considering metadata like another "facet" of the data. -- -- Bertrand Delacr�taz, www.codeconsult.ch -- web technologies consultant - OO, Java, XML, C++
