RE: [xml-dev] XML and the Relational Model (was Re: [xml-dev] A standard approach to glueing together ...)

[Jeff Lowery <Jeff.Lowery@creo.com>]

> I'd propose that an XPath statement in general can be considered the XML
> equivalent to a view...

I'd propose that an XML document can be considered equivalent to a
domain-optimized storage format for a relational front-end, provided that
said document is information rich enough to fully support relational
operations.

Manipulating the XML by other (non-relational) means is equivalent to taking
shortcuts, with risks of violating data integrity.

Shoot, if I want to change a value in a Java properties file, I may just
whip open Notepad.exe and edit it.  The proper, integrity-checked way to do
it is to write a little program that opens the property file, reads the
properties into a Property class, call the setProperty() method with the
appropriate parameters, and check for any exceptions.  Then serialize it
back out again. That's more formal and correct, and just plain silly.

It's also just plain silly for relational purists to proclaim "NO! You can't
touch the storage model! You go through the database front-end!".  Of
course, we never do that with database backends because the storage format
convoluted that we wouldn't dare try to poke bits into it ourselves. But
that's not the way XML is.

To embellish Tom Passin's rant:

How about a formal description of the characteristics required for an XML
document to fully support relational operations at the front end? Murali's
and others' work on XML normalization is getting us closer to that end.

Will we all then start writing XML documents that conform to characteristics
outlined above?  Hell no!  Not many of us, anyway.  Careful shortcuts are a
perfectly valid pragma, with plenty of empirical support.  It's the sloppy
shortcuts that cause us problems.  XML allows for both. Choose wisely.