notes

The terminology is confusing also, because there is a Reader and there
is a source, but the source is an Input in terms of class. The main
reader is the standalone reader, although there are others.

We also have to consider streams, etc. We probably out to push the IO
on to the client since javascript has no built-in IO, really.

Is there a point to having the reader/writer abstraction? Should it redefined?

publish - reader.read(this.source, this.parser, this.settings, (error, doc) => {
   applyTransforms()
   output = writer.write(doc, this.destination)
   output.writer.assembleParts();

reader.read (source, parser, settings, cb):
    this.source.read((error, data) =>
       this.input = data;
       this.parse(); -> sets this.document
                        there is no reason to not simply return the document
			from parse!
       cb(error, this.document);
       

reader.read can become a single callback, screw the class!
maybe accept the class?

Any I/O will involve asynchronicity.
This is a main difference.