> ...For segmentation purposes, shouldn't "noun"
> and "Group of Islands" be in different <trans-unit>
> elements?
They actually should not be at all in any <trans-unit> I think--at least not
as content--since they are not part of the extracted text.
That's why I would be concern if extensions elements in <source> and
<target> were to introduce additional "structure" (I'm still not sure if
it's the right word). Having chunks of content not belonging to the actual
text of the <source> or <target> would seem to lead to much more work for
handling the text itself.
Christian's example of <olf:term> illustrates the way that would probably be
preferable when one needs to associate a lot of data with a specific part of
the <source>/<target> text.
Maybe there are scenarios where "embedded" content is the only way... I'm
trying to think of any, but come empty so far.
-yves
