Data streams and schema use and identification

If a data stream is supposed to be validated with a schema or processed 
against one, is there anything REQUIRED to be in that stream that would 
identify it with schema processing?

I know that some tools will allow you to take a well-formed data stream and 
via some outside mechanism identify that it should be validated with a DTD 
or Schema without reference (or maybe counter) to anything in the data stream.

I believe that a DTD world requires a DOCTYPE identification (preferably 
with a SYSTEM and even better a PUBLIC identifier) and that a schema world 
requires one or more of the xsi attributes to make the connection to the 
schema to be compatible with the specifications.

What tools do to work around issues of schema and DTD management/location 
(catalogs or feeding a value in some other manner) should not necessarily 
be considered standard or at least good practice.

comments?

..dan
---------------------------------------------------------------------------
Danny Vint
http://www.dvint.com