Supported
---------
Note: any combination of these is possible
- selecting root element: /
- selecting all elements: //
- selecting all enclosed elements: *
- selecting attributes: @
- selecting elements using attributes as keys: [@x]
- selecting element using attribute and its value as key: [@x='y']
- selecting descendents: descendent:: , descendent-or-self::
- selecting ancestors: ancestor:: , ancestor or self::
- selecting siblings: following-sibling:: , preceding-sibling::
- selecting following and preceding elements: following:: , preceding::

Probable support
----------------
- using the 'and' operator ('|' in xpath) to combine rules
- selecting first element: [1]
- selecting last element: [last()]
- selecting elements that don't have attributes [not(@*)]


Not needed
------------
- selecting elements with x number of children elements: count()
- selecting element using an attribute value and ignoring leading spaces
in the value: normalise-space()
- selecting elements according to length of element name
- selecting elements in a certain position in a document (using
mathematical expressions).


- A watcher wishes to get the tuple that has a <class> element with
value "IM". The watcher has to either explicitly list all the elements
that can appear in a tuple that it wants to receive, or alternatively,
it tells the PA to deliver the tuple that has element <class>IM</class>
with all the other elements that appear in that tuple. Its like using a
key for selecting a whole tuple. The latter can be achieved using the
following expression:

//*[rpid:class="IM"]/descendant-or-self::*


This actually selects the whole tuple only. if the watcher wants to
select the tuple and the root element <presence>, you would do something
like:

//*[rpid:class="IM"]/descendant-or-self::*  |
//*[rpids:label="im"]/ancestor::*


- The following is an example of how you can ask a PA to include all
tuples that are not of <class> IM:

//*[rpid:class="IM"]/following::* | //*[rpid:class="IM"]/preceding::*

- A watcher asks the PA to deliver a tuple that has a <basic> status of
"closed", but it only wants the elements that are not <status>

//*[basic="closed"]/following-sibling::* |
//*[basic="closed"]/preceding-sibling::* | //*[basic="closed"]/ancestor::tuple


vnd and prs URI schemes
http://www.ietf.org/internet-drafts/draft-king-vnd-urlscheme-03.txt