Cocoon Pipeline Processing Internals -- Some Questions

[Stuart]

Hi,

I am relatively new to Cocoon and have some quick questions as to how
it processes its pipelines:

1. I assume that each stage in a pipeline must complete fully before
the next phase is invoked, right? In other words, it is a serial,
step-by-step pipeline, not one that ever works in parallel a la classic
reader-writer processes sharing a buffer.

2. Is the XML that passes through the various pipeline stages reparsed
from text at every stage, or once the generator parses the original XML
into a SAX data structure, is the XML passed to the various
transformers as a SAX data structure?

I ask these questions, particularly the latter, to try to understand
the performance implications of using Cocoon generally and its various
component specifically.

Thank you very much to anyone who can enlighten me, either via comments
or links to enlightening writings, on these basic topics.

Stuart Hoffman
Vienna, VA