I am in the process of deleting what was chapter 32 (CR) in P4 from
P5, and inserting SO W 08 into what was in P4 chapter 14 (SA). I have
some questions that need quick answers.

I am pretty sure we originally intended to allow something like

   <ptr cref="14.2" decls="#madeUp"/>

with either

  <refsDecl xml:id="madeUp">
    <fragmentPattern re="([0-9]+)\.([0-9]+)"
                     pat="xpath(//div[$1]//egXML[$2])">
      <refDesc>a canonical reference to an XML example is the
        (sequential) division number followed by the (sequential)
        <gi>egXML</gi> number within that division.</refDesc>
  </refsDecl>

or

  <refsDecl xml:id="madeUp">
    <fragmentPattern re="([0-9]+)\.([0-9]+)"
                     pat="#xpath(//div[$1]//egXML[$2])"/>
  ...

(the only difference is the "#" in front of "xpath") to permit
software to generate the equivalent of
   target="#xpath(//div[14]//egXML[2])"

Two questions come up.

1) Which to use, "#" included in the pat= value or not? More on this
   below.

2) Could someone remind me why we used a special element, <refDesc>,
   for the prose description, rather than either using %paraContent;
   or ( p+ ) as the content of <fragmentPattern>? Right now I can't
   think of a reason to create a new, special element over one of
   these other methods.

so far so good, although we need to know which one. I vote for the
2nd ("#" included in the pag= value)


Furthermore, as (I think) the current thinking goes, something like

  The <ref cref="05:02:id2822130" decls="#rng">element pattern</ref>

when teamed up with something like

  <refsDecl xml:id="rng">
    <fragmentPattern re="([0-9][0-9]):([0-9][0-9]):([A-Za-z0-9.-]+)"
                     pat="http://books.xmlschemata.org/relaxng/ch$1s$2.html#$3">
      <refDesc>a canonical RelaxNG reference of the form 2-digit
         number, colon, 2-digit number, colon, identifier is for a specific
         spot in a particular chapter</refDesc>
    </fragmentPattern>
    <fragmentPattern re="([0-9][0-9]):([0-9][0-9])"
                     pat="http://books.xmlschemata.org/relaxng/ch$1s$2.html">
      <refDesc>a canonical RelaxNG reference of the form 2-digit
         number, colon, 2-digit number is for an entire
         section</refDesc>
    </fragmentPattern>
  </refsDecl>

would (permit software to) produce the equivalent of

   target="http://books.xmlschemata.org/relaxng/ch05s02.html#id2822130"

In the #madeUp example, the net result of cref=, re=, and pat= is a
fragment identifier. In the #rng example, the net result is a
complete absolute URI. This brings up two more questions.

3) If a complete URI, as opposed to just a fragment ID, can be
   generated, shouldn't the name be something other than
   <fragmentPattern>?

4) Are other combinations permissable? (I think the answer was "yes",
   but don't recall for sure.) I.e.,
     cref="duck/bill"
     re="(.+)/(.+)"
     pat="$1.html#$2"
     xml:base="http://some.addr.com/animals/"
   resulting in http://some.addr.com/animals/duck.html#bill.

If the answer to (4) is "yes", then I think it's a given that the answer
to (1) has to be that the "#" must be included in the value of pat=.