On Thu, 23 Oct 2003, Christopher Kings-Lynne wrote:
> > You couldn't have said better what I meant.
> > I store the xml already parsed. You can navigate right along. To the
> > parent, the previous, the next elemnt or the first or last child.
>
> Which is the whole point of indexing it...
not quite. Indexing can solve some of the problems, not all.
1) You have to update the index every time you modify the data. My custom
format serves as an index for some queries.
2) The page format is designed in such a way that modifications
(insertion, deletion) are as fast as the original parsing. I'm not sure
how that stacks up to modifying data in a column. I guess it depens on the
strategy to store very large strings in columns.
> >>I use XML a lot for all sorts of purposes, but it is appropriate for
> >>data transfer rather than data storage, IMNSHO.
> >
> > Right now, you're quite right. But I want to change that.
>
> No point, it's a data exchange format, it's not usefull for data storage.
Well, neither one is a data exchange format only or a data storage format
only.
Rather, the difference is that relations are designed to store structued
data while xml is desinged to store semi-structued (not so regular) data.
Which is better suited for data exchange is a matter of convention (where
xml seems to be good), while efficiency and other features of an
implementation determine, which one is suited for data storage.
If your point is that currently xml is not suited for storage, because
there are more efficent RDBMS than xml databases, I agree. Otherwise, I
don't see your point.
--
Gregor Zeitlinger
gregor@zeitlinger.de