Lists Home |
Date Index |
3/15/2002 9:22:36 AM, "Bullard, Claude L (Len)" <firstname.lastname@example.org> wrote:
>Maybe I am just getting foggy in my dotage, but do we
>need Xlinks if we use relational dbs and if so, for what
>other than perhaps a convenient transport representation?
That's an awfully good point. As anyone who knows who I work for
can imagine, I think the "native" XML approach has a lot of
good use cases. Nevertheless, maintaining relationships (and referential
integrity) is simply not something that XML is very good at.
Linkbases are "better" than hardcoded links in this respect, but
still fall afoul of C.J. Date's frequent rants against using
"pointers" to relate pieces of data to one another. Anything displayed
via the web will ultimate resolve to hyperlinks, but that doesn't
imply that they should be stored and processed as "pointers."
I think the most significant advantage that XPath 2 / XQuery will
offer is "joins" to handle relationships that are implied
by values rather than demanded by pointers. This is exactly
Date's point -- find relationships that exist at runtime with joins
rather than hard-code them with links/pointers.
So, if you're using an RDBMS today, I agree that XLink doesn't
offer much. In XQuery-aware XML databases that are being
prototyped today, that may also be true ... I'm still sorting out
in my own mind how XLink/Topic Maps and XQuery fit together.
I suspect that some future Topic Maps powered by XPath 2 will be able to
do some interesting things that would be very difficult in
an RDBMS ...
- From: "Bullard, Claude L (Len)" <email@example.com>