Lists Home |
Date Index |
----- Original Message -----
From: "Dare Obasanjo" <firstname.lastname@example.org>
> Doesn't anyone else see something wrong with THREE different languages
> significantly differing syntax that have roughly the same functionality
> (querying) with yet NONE have update or delete semantic? However some
> think that the time has come for XML databases? Interesting..
> PS: IMHO, until someone comes up with a language with simple
> SELECT-DELETE-UPDATE semantics, XML databases will be the
> OODBMSs of the new millenium.
That's the language I'm trying to design right now,
it looks possible. I would be glad to discuss
the problems with anybody who may be interested -
it is really interesting stuff and it seems that no wheel
In my oppinion, to design the language like that,
one should start with refactoring XPath syntax, so
that it will become 'XPointer', instead of 'mighty XSelector'.
To make it handy, I think that the number of axes should
be minimized and the XML Node (Atom) should be
slightly changed to give more air to unordered properties.
When XPath-- would be ready, that would
allow some tricks, comparable to SQLs tricks.
By the way, SQL is kinda inconvinient, because it
has no 'UPDATE-OR-INSERT', so that common
atomic action is usually implemented with 2 queries
( first select and if nothing exists - insert, otherwise - update ),
using two queries for one atomic operation implies
transactions e t.c.
> I remember being introduced to XML and thinking that the concepts
> behind relational databases were more complex than those behind the
> hierarchical structures that encompass XML, amazingly enough the W3C has
> proved me wrong by producing increasingly complex languages that
> deal with handling XML in databases yet have much less functionality than
> simple language like SQL. *sigh*
Well... I should say that I think that SQL is not actually simple.
In SQL we have operations ( select - insert - update ),
but also we have schemas ( create - drop - remove table )
and constraints ( that's a reall mess, actually )
My experience with PXDB shows that
balancing these three things is *really* hard, because,
for example, schemas overlap with constraints and
I had a hard time just trying to understand on what level
to put some particular constraint, like 'this property has
to be unique'. This is actually very interesting topic.
Unfortunately, to produce handy bidirectional XPath
one should produce 'really handy' Node model and my
opinion is that things like minimizing the importance of
properties order is the way to go. XML Chunks v 1.0
looks like a first step in that direction, but it is sure
not the last step.
I think that when XML would get a nice Atom, the rest
would be simpler, than it is now.