On Fri, 13 Feb 2004 09:04:36 -0500, David Sewell wrote: > So that means that to return "letter" nodes across those publications > we'd need to use XPath like > > (//div1|//div2|//div3|//div4)[@type="letter"] > > which is ugly as sin. (And what if a new publication goes to div5?) > Or we could be lazy and wildcard: > > //*[@type="letter"] > > which could be very bad if anything other than a <divN> element > had a @type of "letter". Does it matter that an expression is "ugly" if it works? Another possibiliy for matching any possible numbered div when it is a letter, while avoiding letters that are encoded as something else: //*[starts-with(name(),'div')][@type='letter'] // Gregory Murphy. Isopaleocopria.