RE: Bob's letter on V5R4 RPG stuff -- RPG400-L

Hi Jon,

<snip>
>> Does this XML implementation support DTDs or XML schemas?

No - not yet.  It makes the assumption that the XML is well-formed and
valid.  Although if there are missing/extra elements/attributes you can deal
with it.
</snip>

That's a big assumption. How do we check for well-formedness prior toprocessing this document? This is a BIG issue with xml processing. If youuse the simple SAX parsing model and you are not using transaction controlsuch as commitment control then you have a real headache if your doc goesbang halfway through processing it and you have already written data totables. DOM is different, as it builds the object model in memory and willfind errors before you get a handle to it. But this is expensive for simpleread-only processing.



<snip>
>> And imho any parser should work on those DTDs/XML schemas.

Not sure why you say that.  I have never noticed any that do - but then I
wasn't looking.
</snip>

The XML Toolkit provided by IBM to run on the iSeries contains the serviceprogram QXML4PR531 in library QXMLTOOLS. This service program is designed toallow procedural languages like RPG to use the DOM and SAX parsers providedwithin the toolkit. I know for certain that the SAX parser in there supportsfull schema validation because we use it in our RPG programs.

I would say, though, that we only have full schema validation turned on whendeveloping/testing/debugging. The beauty of validating schemas is verysimple - it defines the interface. When we define an interface to a newsystem we also define a validating schema. This definition should be allthat matters to both parties with regard to the architectural side of theinterface xml. This means I can test my xml generation and as long as thedata I will pass validates against the schema I know I'm ready to go. Thisis invaluable when you are testing in a sandbox prior to hooking up with theother party. It would be a nightmare if I couldn't get going with mydevelopment without constant reference to what is happening with thedevelopers at the other side of the interface.

Validating schemas provide a reference point - if something goes bang on theother side, we can validate the xml to see if it is "valid". If not, we fixthe xml generation program, if it does we get them to fix their xmldecomposition program. In the rare occasions where the doc is "valid" butincorrectly defined we get together and re-engineer the schema. This thenbecomes the new "point of reference" for the interface guys.


Life is MUCH easier with validating schemas. :-)

Cheers

Larry Ducie