RFE: mode attribute for XWEB fragments

I would like to propose that a "mode" attribute be added to the fragment
element in the XWEB literate programming schema, with the following
semantics: if no output mode is specified by a user (for example, as an
parameter to an XML transform), then the "mode" attribute of the "top"
fragment will provide that information to the application.  I have
attached a patch to the CVS version of the ldocbook.dtd file which
implements this simple change.

By specifying the default for the "mode" attribute to be "text", I have
made a subtle design decision which I feel warrants a sentence or two of
discussion.  Text output can be provided even if XML tags exist in a
subtree, whereas a default of "xml" may not always produce well-formed
XML (if the value should have been text).  In addition, I believe
processors ought to recursively output all text children of every
subtree when processing using "text" mode - this would enable authors
to mark up the content of fragments and then have tags therein ignored
on output with a mode of "text".

In terms of regression, I anticipate this will not break anything, as
previously every processor (tangle, weave, etc.) explicitly specified a
processing mode.

If this were to be accepted, I would be happy to update the
documentation - is there CVS access to the "Literate Programming in XML"
document anywhere?

Thank you for your consideration,

    John L. Clark