On Fri, 23 Jul 2004, Martijn Tonies wrote:

> > > Native XML storage? What's that supposed to mean? Rubbish!
> >
> > Yes, it's a horribly misleading and confusing term. I think that when
> > people say they want "native XML", what they really want is an XML column
> > type that supports DOM, XPATH, or some other XML-specific operations.
> > This is not a bad idea, IHMO.
>
> Not a bad idea? Why not? What exactly does XML mean then? What
> would you expect the engine to do with it? Validate it? Ask for a specific
> node?

Mean? What does a VARCHAR mean? What does a DATETIME mean? It's just a
domain. The meaning comes from the predicate that backs the relation.
I see where you're going with this, and I don't necessarily disagree; in
general, I don't see much that XML has to offer over relational aside from
its usefulness as a packaging / transport mechanism for data. Why does XML
need to be in the database? I don't know, but from a theoretical
standpoint, it's just another type with a set of well-defined (albeit very
generic and string-oriented) operations.

I'd expect the engine to do at least the following:

- Ensure syntactical correctness
- Validate against an XML Schema or DTD
- Perform queries on subelements
- Perform modifications to the document model

> XML is human readable crap used to exchange data between systems
> that don't need human readable crap with too much bandwidth to spend :-)

I disagree. It's not always human readable. =)

> > "Native" seems to imply that XML is actually saved to disk as XML, which
> > seems a) irrelevant, since data independence dictates that we should not
> > care how it's saved to disk, and b) undesirable, since most of the things
> > people would want to do with XML would perform much better with a parsed
> > DOM, not straight text. But of course, this doesn't stop marketing
> > departments from using the term to promote their clearly superior,
> > next-generation technology. ;)
>
> Ho hum... talking about a step back :-)

Well, I think it depends on your requirements. Maybe XML is a
self-fulfilling prophecy, but it seems to me that if I were given the task
of managing a bunch of XML files, and I didn't have to do deep
modifications to them, I wouldn't bother converting between XML and
relations. I'd just find some way of managing them as XML. Of course,
whether the XML was necessary in the first place or not ought to be
debated, but if it's a fixed requirement, I'd probably start searching for
a database that understands the syntax of XML.

> > The ability to manipulate XML data through SQL seems more useful than the
> > ability to dump relational data as XML. Some DBMSes provide functions for
> > projecting data into XML, but since there's no standard mapping from
> > relations->XML, every DBMS may conceivably do it differently, and none of
> > these variations may be what the application already needs. For instance,
> > one common usage of XML is to send structured data to Flash, but if you
> > have no control over the format required by the Flash movie, you're back
> > to writing procedural code anyway, unless your DBMS's mapping function is
> > highly configurable (and it isn't, typically). Not a big deal, IMO.
>
> That's the problem... "XML" by itself isn't defined at all...

I have a feeling you're talking about semantics again here, where I'm
simply talking about types. XML as a type is just a recursive tree
structure that happens to look like a bunch of nested tags, just like a
DATETIME is a record structure that happens to look like a formatted
string. Or, at least, that's one potential viewpoint.

> btw, I too use XML for some things - for example, configuration.
> It's quite useful in those cases...

Perhaps. I'm not a big fan of XML configuration files. They're too wordy
for my tastes. However, I think it's good for interop, since a lot of
languages and platforms have support for it. Whether or not there's any
point to persisting XML in a database really depends on the requirements
of the application, I think.

--
.:[ dave benjamin: ramen/[sp00] -:- spoomusic.com -:- ramenfest.com ]:.

"When the country is confused and in chaos, information scientists appear."
Librarian's Lao Tzu: http://www.geocities.com/onelibrarian.geo/lao_tzu.html