[Rets-dev] Why XML?

[Matt Lavallee]

On Friday, August 03, 2007 at 10:33 AM, Dave Sullivan wrote:

		#2 does not entirely cover the scope and reality
		that XML is also restrainable.


>	But why should it be restrained?  To make it "easier" to process?

To make the data predictable between systems.

>	Well ... City should use the Standard; and the Standard should not
be so rigidly defined (as to length, etc.) to force users to define
separate, local elements in order to accommodate existing content.  In our
semantic-centric approach, the standard name should be used to convey
semantic content with non-semantic attributes used to convey
computer-related aspects (data type, max length, etc.) if this information
is found to be necessary or useful.

REIL redefined City in their system to return keys (integers) instead of the
full city name.  I understand the argument from a bandwidth/optimization
point of view, but it breaks the contract of client expectations from a
standards point of view.  Permitting this in the new standard is not OK with
me.

>	I suggest that the perceived "failure" of StandardNames is that they
were not used and the "standard" was overly rigid as to format, etc.

I was more pointing to the fact that implementors will use the path of least
resistance to be "compliant".  If compliance allows circumventing the
standard, then we've achieved nothing better than RETS 1.	

>	Matt - I agree that a DTD can convey content for a well-known
transaction; just as X.12 can convey it.  It is conveying content in the
absence of agreement on the transaction that is the trick.  You seem to
believe that detailed agreement on the exact content of a message is
required prior to communication; while I believe that the semantic approach
allows communication without prior agreement.  

As I said, I am not opposed to custom schema for custom payloads.
Obviously, there is a need for semantics in that model.  However, the
standard payloads have a well-considered, rich data format, already complete
with annotations that can still be enhanced locally without affecting the
data model itself.

-Matt