You might try putting two imageobject elements inside a mediaobject
inside your screenshot, and assign an appropriate role attribute to
each imageobject, such as "html" or "pdf". The imagedata element
in each imageobject could have different attributes for scaling the
Then you could use the profiling stylesheet as a preprocessor to
select which imageobject is to be used before passing it to the
html or fo stylesheet. It is a bit cumbersome, but it should work.
Interesting. I'll try this tomorrow.
I couldn't use the role attribute, because I was already using
I used the arch attribute instead, so the images look like this: