> I would propose for the request to have space or plus sign separated
> recordSchema's. For a generic wrapper there are lots of 
> possibilities. I
> give some suggestions but there may be better ones. I also like Bill's
> RDF suggestion, but only if the server is allowed to choose 
> between RDF
> and a generic wrapper.

Firstly, I don't think we need to change anything in the response - we
already have wrapper schemas such as METS and MPEG21 DIDL which do the
job already.

In the request, I was thinking of having as XML structure such as:

<compoundRecordSchema>
  <containerSchema>url_to_METS</containerSchema>
  <containedSchemas>
    <recordSchema>url_to_OpenOffice</recordSchema>
    <recordSchema>url_to_MODS</recordSchema>
    <recordSchema>url_to_REC</recordSchema>
  </containedSchemas>
</compoundRecordSchema>

This could be substituted (if required) for recordSchema in the request
i.e. the relevant bit of the request schema becomes

<xsl:choice>
  <xsl:element ref="recordSchema"/>
  <xsl:element ref="compoundRecordSchema"/>
</xsl:choice>

compoundRecordSchema could also be substituted for recordSchema within
itself for where you have compound schemas within compound schemas eg

<compoundRecordSchema>
  <containerSchema>url_to_METS</containerSchema>
  <containedSchemas>
    <recordSchema>url_to_OpenOffice</recordSchema>
    <compoundRecordSchema>
      <containerSchema>url_to_MPEG21_DIDL</containerSchema>
      <containedSchemas>
        <recordSchema>url_to_DC</recordSchema>
        <recordSchema>url_to_MARC</recordSchema>
        <recordSchema>url_to_MODS</recordSchema>
      </containedSchemas>
    </compoundRecordSchema>
    <recordSchema>url_to_REC</recordSchema>
  </containedSchemas>
</compoundRecordSchema>

So there's ouor solution, now we just wait until we get a requirement
;-)

Matthew