After feeding the DTD, we want a DTD memory model as close to DOM as possible
which will present the DTD as a tree format so it can treverse. This will
generate a DOM like object which is not text, but, rather a memory object
so that we can do real-time comparison with DMS. This concept is very similar
to XML DOM model.

In a similar fashion, we may want XML memory model as opposed to a flat text
file. The main benefit of an XML memory model seems to be in XML chunking.