Cart
[Home] [By Thread] [By Date] [Recent Entries]
I am glad to announce three works related to XML: I. Constraint Specification for XML, available at http://www.cs.ucla.edu/~mani/webdb2003/constraints.pdf The main things to note here are: a uniform framework by which we can compare different constraint specification schemes. They all specify the keys in terms of (rel, sel, field), this means for an element that satisfies rel, the key for all elements that satisfy sel is given by the set of path expressions in field. Different proposals use types or path expressions for rel and sel, whereas field is always a set of path expressions. For example, XSD uses type for rel, and path expression for sel. Suppose XSD says something like: Library -> (Book (sel: Author, key:name), Book, Paper) then we can actually consider it as Library -> (Book1, Book2, Paper) and then the key (Book1, Author, name) XML Schema actually uses path expression for selector. Besides this, our main results include (a) A new proposal where we use type for both relative and selector axes - I argue that this seems more appropriate. (b) Given the above assumption, we can prove that all we need is only path expression leading to values in field. We do not need path expressions leading to nodes.. To sum up, our key specification is (rel, sel, field) where rel and sel are types, and field is set of path expressions. Our foreign key specification is similar, and we also provide additional capabilities for identifying target types for IDREF and IDREFS attributes. Our work is again at a theoretical level, which can be translated to practice. II. Normalization theory for XML, available at http://www.cs.ucla.edu/~mani/webdb2003/normal.pdf This talks about redundancies and how to understand them, and how we can eliminate them, if we want to. Redundancies should be well-understood, as they can cause problems with respect to updates etc. Normalization theory is one field where relational DBs have done lot of excellent work, and it is very good that we are able to study XML normalization very similar to relational normalization. III. Another work under progress is how to model real world application scenarios in XML, this is similar to ORM etc. We study this from a DB perspective (almost universal) - that is we want to model classes and relationships between them.. This is work in progress and most of the current status can be found at www.cs.ucla.edu/~mani/er03/modeling.pdf I would be very glad to receive any comments on these works.. thanks and regards - murali.
|
