[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

"Whither thou goest, XQL?"



"Difficult to predict, I suppose.  Not sure if it's relevant, but XSL
Transformations certainly changed significantly during their
development.  There is probably nothing to prevent XQL changing
radically either.

"that future W3C XML query language (XQL ? will be based on XPath,"

I generally concur with both Steve and Zvi.  There already are XPath
implementations in Java and I think there are good reasons (if
we want to get a head start on it) to investigate the ways that XPath
does
basic XML node selection with a view to producing a low-level API for
Ozone.

(I will try to find the link to the XPath implementations and post it
here later.)

Now, OTOH:  once a query algebra has been developed for XQL:  A whole
different problem could emerge: query re-writes for optimization not of
the access plan, but of the query itself, first.  In any "network" of
nodes, there may be more than one path to a node:  optimization involves
computing the fastest path.  If the user specifies a particular path, an
optomizer converts it to the fastest equivalent path.  (A whole subject
in itself!)

Yes, Falko, GMD is *definitely* a "squeaking wheel" in the working group

and they are arguing for a sound, unambiguous basic algebra as the
foundations for a syntax so that being able to express "bad" or
inefficient queries in XQL will be impossible.

I believe that GMD will be pushing strongly for a Datalog-type algebra
with Datalog-type selection operators (much different syntax, however).

INRIA is also doing some squeaking and
INRIA also has a strong background in logic and languages---so (all of a
sudden) there is alot of brain-power pushing for change.

Anybody who has experienced SQL wants that sort of mess not to happen
again.
(SQL isn't sound, efficient, unambiguous, or easy-to-understand: "left
outer joins", anyone?)  The original authors of the XQL appear to be
willing to bend and are
being overcome by "no" votes, now.

Right now, (as I mentioned in an earlier message), I would spend more
time getting
familiar with XPath than the original XQL.  Even the original author of
the XQL syntax (Robie) has made kindly comments recently about some
other syntaxes for an XQL.  If you want to understand how logical
selection operators work, read up on Datalog.

As far as an XQL "update" feature:   I haven't seen any discussion for
it.  Actually, I am a great deal more concerned with the whole area of
fragments and super-documents, links, and so on.  I think even this is
off-limits for
current working group right now lest they never get anything out the
door.

The big problem right now: they could change course radically, at any
time, and we won't get a single word of notice about, because they are
really and truly bound to strict non-disclosures on both the substance
of the meetings *and* any private communications related to same.

BTW, the *one* thing I haven't heard anybody on the comittee say a
negative word about is the use of Unicode, the general XML naming
scheme, XML namespaces, etc.

Also, I am hoping that by the time a new XML query language does emerge
that we might get a new implementation of Antlr that handles Unicode
correctly---this would make a syntax parser a whole lot easier to
maintain for the long run. BTW,  I am communicating regularly with the
OpenXQL org on this.  They also appear to be in a "wait and see" mode.

I see *alot* of  good, sensible work coming from this group.  They may
actually end up doing it right!

Best Wishes,

Ann Tecklenburg