In the documentation from the URL above (version 1.2), section 7.1 details the
directory structure, including a description of the node comparison method for
buiding the red-black tree.

The description states in the second bullet point:
If both names have the same length, they are compared character by character
(case sensitive).

I believe that the character by character comparison in the MS COM
implementation is actually case insensitive, so the bullet point should read:
If both names have the same length, they are compared character by character
(case insensitive).

The evidence for this belief came from writing a pure Python implementation of
the file format based on the document. I found using some nonsense test data
that after creating some streams in the root storage, reading the file using the
MS APIs would return a COM error with message 'The docfile has been corrupted'

I double checked using the same nonsense names from the MS APIs, and found that
the data was sorted differently in the tree.

eg: creating file from MS APIs, the in-order sequence of stream names was as
follows:

'Abbdab', 'doodah', 'Tester', 'thingy', 'Wotsit'

Whereas from a case-sensitive comparison, the sequence was:

'Abbdab', 'Tester', 'Wotsit', 'doodah', 'thingy'

Otherwise, thanks very much for an incredibly useful and well set out document,
Alan