[
Lists Home |
Date Index |
Thread Index
]
From: John Cowan [mailto:jcowan@reutershealth.com]
Bullard, Claude L (Len) wrote:
>> 1. Laissez-faire: send only the message. [...]
>1a. Send the schema out of band, just once
>(or only when necessary). Receiver checks the message
>against *his* copy of the schema, which represents
>his current understanding. This is what I do
>in current XML publishing: I claim the message
>conforms to a separately published DTD (available
>on the Web), but it does not contain a DOCTYPE
>declaration.
Yes, and only verify on error, criticality level,
or opportunity. Trust but verify if suspicious.
Good catch.
>> 2. Schema/DTD travels with message. [...]
>Really, the DTD part of a document is a lot like a checksum:
>it guarantees self-consistency, not consistency
>with anything else.
Yes. I almost used checksum in the description but
didn't want to start an argument about **types**. On
the other hand, the way I should have written that is
to say "You send the test because you
absolutely want the receiver to understand this
message exactly as YOU understand it at this point."
>Whenever I get an SGML feed,
>I always worry that the new document type is
>different from the old document type in a way
>hidden by the packaged DTD, since there is a new
>DTD with each message."
Yes. See above. The presence of the DTD could
be construed to be a request to verify that we
still have the same understanding as before, or
that we may be about to re-negotiate, or a new
negotiation was completed and successful.
>> 3. Ask the Web: use RDF or some other expert system
>> what is needed. Isn't this sort of a dictionary? It
>> works as long as you own or accept the ontology of
>> others. This is Trust and Verify. Advantages?
>This is the same as your case 4 (Java), but using a
>declarative language rather than a procedural one.
Ok, but wouldn't inferencing offer the capability to
test different assertions? Not possible with Java
right?
len
|
