RE: Question regarding order in XPath

> 
> However, I am not sure why we should always try to translate // to
> something like descendant-or-self::node()/child; why not 
> define semantics
> for p//text () as -- all nodes in the document in document 
> order with 2
> predicates -- check if they are descendants of p, check if 
> they are text
> nodes.. then we will not switch off the computer !!??

Here are two useful constructs that would not work (or would not produce the
XPath 1.0 results) with this definition:

p//@code

p//item[1]

There are other constructs that are less useful, but still allowed by XPath
1.0, such as

p//..

Michael Kay