MicroXPath proposal

[John Cowan <johnwcowan@gmail.com>]

I've been thinking a lot about MicroXML, MicroXSD, MicroRNG, etc., and
I thought I'd take a stab at a MicroXPath.  This is intended to be,
like the others, the simplest thing that could possibly work.
However, I'm not a frequent XPath user, so I may have put together
something that is *too* simple to work.  So I'm posting here for
comments.  (I'm using this address because xml-dev drops posting from
my permanent cowan@ccil.org address on the floor, for some reason).

One principle is that since elements are the only kind of node in the
MicroXPath data model, XPath expressions return only lists of
elements.  Another is that 100% compatibility with XPath 1.0 isn't
essential, since most XPath expressions are fairly transient compared
to XML documents or schemas.

So here's what I've got.  The only kind of expression is a location
path.  Location paths can be absolute or relative, and their steps can
be separated by either / or //.  A path step is an element name, star,
dot, or dot-dot.

Only one predicate is allowed per step, and it has to take one of
these eight forms:

1)      a literal number

2)      last()

3)      location-path="string"

4)      @name="string"

5)      location-path/@name="string"

6)      @@name="string"

7)      location-path/@@name="string"

8)      function(.)

@@ means "inheritable attribute", and is provided for brevity and
convenience.  In form 8, the function name is provided externally to
XPath (there are no built-in functions except last()), and accepts the
current element node and returns a boolean value.  This is the
programmatic escape from MicroXPath that allows complex predicates.

What's missing?  What can I take out?  Comments solicited.

--
John Cowan       http://www.ccil.org/~cowan        <cowan@ccil.org>
       You tollerday donsk?  N.  You tolkatiff scowegian?  Nn.
       You spigotty anglease?  Nnn.  You phonio saxo?  Nnnn.
               Clear all so!  `Tis a Jute.... (Finnegans Wake 16.5)