Very simple? Text processing from document data

Hi all,

I'm having a mental block today. As an example, I have:

<p>Some body <b>text</b> some more</p>

And I want to replace the word 'body' only in the text of <p> but ignore
text in <b> so I've been looking at something like this:

<xsl:template match="p">
   <p>
	<xsl:value-of select="substring-before(./text(), 'body')"/>
	<xsl:value-of select="'replace_body'"/>
	<xsl:value-of select="substring-after(./text(), 'body')"/>
   </p>
</xsl:template>

How do I separately select #text before, between and after an arbitrary
number of interruptions (could be tables, images etc) and process those
interruptions in document order?

Thanks,
Richard