In the manpage for the xsxp package the example is given :
xsxp::fetch $pxml {html body p#4 b} %PCDATA
This should be corrected to:
xsxp::fetch $pxml {body p#4 b} %PCDATA
fetch does not use the root node ( i.e. html ) by name and the fetch function purposely steps over it. In fact the only way to return any information about the root node is to specify {} as your path.
Having looked at the code one could also update the manpage to say that xsxp::fetch accepts paths in both list and '/' delimited format (e.g { body p#4 b } and /body/p#4/b )