Thank you very much for the answer. There is a slight misunderstanding
on what I want to do with the breaks. For example, my example 2:

> Example 2:
> <xml>
>   <p>Aaaaaa</p>
>   <break/>
>   <p>Bbbbbb</p>
> </xml>

should be transformed as follows:

<xml>
  <section>
    <p>Aaaaaa</p>
  </section>
  <section>
    <p>Bbbbbb</p>
  </section>
</xml>

In other words, <section> elements should contain everything that
starts and/or finishes with a <break/> element. And therein lies the
rub, since the <break/> element only appears in between "sections", so
it is difficult to programmatically find the start of the first section
and the end of the last one.

The caveat about character encoding suggests that this may not be the
best solution in any case, since several of the files are in Spanish
and therefore contain a very high proportion of non-ASCII characters.
But aside from that, the SGML solution you proposed is interesting -
it's good to know it still has its uses even though I haven't used it
for years (over 10, in fact - boy, I feel old :)

Ariana