An interoperability problem?

[Norman Walsh <ndw@nwalsh.com>]

Imagine two processors, GreedyProc and LazyProc working on this
document:

  <?xml-stylesheet href="http://example.com/a.xsl" type="text/xsl"?>
  <?xml-stylesheet href="http://example.com/b.xsl" type="text/xsl"?>
  <!DOCTYPE doc PUBLIC "x">
  <doc/>

When greedy encounters a stylesheet PI while parsing a document, it
immediately retrieves the stylesheet and compiles it. The parser in
the lazy processor has no special knowledge of the stylesheet PIs, so
it just puts them into the DOM for the document and the processor goes
looking for them there later, if it needs them.

Now let's add an oasis-xml-catalog PI for myCat.xml, where myCat.xml
contains:

  <uri name="http://example.com/a.xsl" uri="file:///stylesheets/aPrime.xsl"/>
  <uri name="http://example.com/b.xsl" uri="file:///stylesheets/bPrime.xsl"/>

For the sake of argument, aPrime and bPrime are very different from a
and b on example.com.

Here's the new document:

  <?xml-stylesheet href="a.xsl" type="text/xsl"?>
  <?oasis-xml-catalog catalog="myCat.xml"?>
  <?xml-stylesheet href="b.xsl" type="text/xsl"?>
  <!DOCTYPE doc PUBLIC "x">
  <doc/>

Now, the greedy processor is going to retrieve these stylesheets:

  http://example.com/a.xsl
  file:///stylesheets/bPrime.xsl

But the lazy processor is going to retrieve:

  file:///stylesheets/aPrime.xsl
  file:///stylesheets/bPrime.xsl

I don't see any way out of this interoperability problem: I don't
think we can say either of these processors is "wrong" in some way and
I don't think we can assert that oasis-xml-catalog PIs have to come
first. Some other spec (say with a PI to control the XML processing
model or something) might legitimately make the same claim.

Thoughts?

                                        Be seeing you,
                                          norm

-- 
Norman.Walsh@Sun.COM   | Wink at small faults; for thou has great
XML Standards Engineer | ones.--Thomas Fuller (II)
Technology Dev. Group  | 
Sun Microsystems, Inc. |