XInclude, images and HTML output

[Dan Sanderson <lists@dansanderson.com>]

Hi all -

The DocBook HTML build process uses the xml:base attributes left behind by 
XInclude to produce file paths to images that are relative to the original 
source locations.  This is a neat idea, but I'm having difficulty figuring 
out best practices to take advantage of this.

In particular, my HTML output has file:/...-style absolute file system 
paths to all my images, which is not useful to me, at least not without 
post-processing the generated HTML.  I notice that my XIncluder is 
producing xml:base locations with a broken prefix of "file:/", as opposed 
to "file://", so img.src.path produces silly values like 
"images/file:/...".  I'm not sure if that bug impacts other things.

What is a typical arrangement involving XInclude that would produce a 
self-contained set of HTML pages (all paths relative to the containing 
directory), with all supporting images in an images/ directory?  What 
should my filerefs look like, and what pre- and post-processing should I 
do to get my image files in the right place?

If there's a way to turn off the XInclude image path resolution feature 
entirely and just pass through my fileref as the img src, I'd be happy 
with that.  But I'm also interested in best practices for taking advantage 
of the feature to copy my images into place, and produce correct paths in 
the HTML.

Thanks for any assistance you can provide!
-- Dan