 |
 |
Re: > From: Peter Graham, Rutgers University Libraries > > Misha's presentation is breathtakingly clear and informing. I must be > missing something. --pg No, you're not. It's a clear illustration of perspecuity revealed by explicit hierarchy. The same information can be represented (er...cobbled up) using simple linear/flat structures that employ private semantics based upon adjacency, etc. But IMHO they will never be as self-documenting as explicit hierarchy. There are indeed problems too with trying to use "containment" *alone* to model important semantic relations, so the XML/SGML notation is no panacea for all the tough problems. Another advantage of the hierarchical notation is that one can provide any datum with an identifier, and reference it, and predicate additional things about it (using attributes). The problem with empty elements (the HTML <meta> element, for example), is that the relations between information objects inside the tag are not easily expressed, and cannot be validated using SGML facilities. One has half a chance with nested elements. [Barely relevant, and only for the patient and truly interested: For those following the XML work, and the recent XML:lang attribute, the 'problem' is evident: the value of the XML:lang attribute governs the element content *and* the CDATA values of all other attributes. Not bad for a default, but you're in trouble (i.e., it gets UGLY fast) when you have to face the possibility that the language of the CDATA in the 'foo' attribute, in this particular tagging instance, is *not* the same as the language of the element content, and likewise not the language of the CDATA commonly applicable to 'foo' and declared as a default. The moral: go with hierarchy all the way to the bottom of your information model; don't multiply private sublanguages in the VALUEs of attribute-value pairs, and don't proliferate the long permuted 'dot.hierarchy' names for attributes (or elements). Limited wisdom, but perhaps worthy of thought: I'm not intimately familiar with your problem domain, so I can't say with confidence. Robin ------------------------------------------------------------------- Robin Cover Email: [log in to unmask] 6634 Sarah Drive Dallas, TX 75236 USA >>> The SGML/XML Web Page <<< Tel: +1 (972) 296-1783 (h) Tel: +1 (972) 708-7346 (w) http://www.sil.org/sgml/sgml.html FAX: +1 (972) 708-7380 ===================================================================