Have you considered implementing serialization of sequences of XdmItems
rather than documents? That way you could preserve the distinctions
that are being lost when you go to a document. If somebody wants to
make a document from your serialized format, they could always construct
it easily enough.
-Mike
On 1/12/2011 7:36 AM, David Carlisle wrote:
> On 12/01/2011 12:23, David Lee wrote:
>
>> My extrapoliation from this is that only atomic and text nodes are
>> concatenated into text nodes, and other node types where allowed are
>> left as
>> a sequence,
>> as part of constructing XDM (not serializing it).
>
>
> When the content of a document or element node is constructed from a
> sequence, you do not get the original sequence ever. atomic items are
> made into text nodes, and adjacent text nodes are concatenated, empty
> text nodes are dropped, attribute nodes either generate an error or
> are moved to the attributes of the element being constructed, and
> other nodes are _copied_ so have a new identity and possibly new
> in-scope namespaces, depending on settings.
>
> David
