It's certainly interesting and instructive: I would never have thought of
doing the metadata in quite that way.
Is it really practical to have quite so much tagging? I know it will
compress
beautifully for transmission over a network, but compressing a gigabyte
file is not something one does just for fun. Using shorter tags (e.g. <c1>,
<c2> for the cells would be a simple fix, but it's still much bigger than a
CSV file.
I would be interested to know why you chose the representation
<row><col1>A</col1><col2>B</col2></row>
rather than
<row col1="A" col2="B"/>
One other minor comment, are you doing anything to preserve white space
in the data content?
Mike Kay