XOM API for Processing XML with Java

[Elliotte Rusty Harold <elharo AT metalab.unc.edu>]

I've been looking at how much a subclass of Serializer should be able to 
do and actually can do. The speciifc context is an effort to write a 
Serializer subclass that outputs canonical XML. Here's what I've discovered:

1. It is possible for subclasses to filter entire nodes as they're 
serialized. That is, it can replace elements, modify text, strip out 
comments, add comments, and so forth.

2. It is not currently possible to change the way data is output at 
sub-node level. For instance, the subclass cannot choose to lexically 
order the attributes as required by canonical XML or decide what does 
and does not go into a CDATA section.

As a result of 2, a CanonicalXMLSerializer would need to basically 
reimplement every method in the superclass, up to the point of 
maintaining its own OutputStrea. There would be essentially no 
implementation reuse, only interface reuse.

Is this a problem? Are there other, plausible use-cases where a subclass 
would want more control of the output? If so, should this be provided by 
publically exposing the TextWriter class or by adding delegating 
protected methods to Serializer such as

protected void writeMarkup(String s)
protected void writePCDATA(String s)

An additional advantage of exposing TextWriter publically is that it 
might make it possible for third parties to add support for additional 
encodings in their own applications (by subclassing TextWriter) without 
having those added to the nu.xom package.

I can see at least three sides to this question. Thoughts and comments 
are appreciated.

--
Elliotte Rusty Harold