[
Lists Home |
Date Index |
Thread Index
]
John Cowan wrote:
> > as SAX, but we felt that the pull model was more approachable
> > for the average developer...
>
> I quite agree: it is superior because of its better fit with
> programmer insight. Files, pipes, database cursors behave in
> just this way: pull to read, push to write.
This has definitely been my experience. I used to be in love with SAX
until I started using .NET's XmlReader. For 9 of 10 tasks, the
pull-model approach is clearly more intuitive and easier to "get right"
for most of the developers I come in contact with. FWIW, now there's
also a JSR (173) in the works for a standard Java pull-model API[1].
Also, it's interesting to note that .NET's XmlWriter feels almost
identical to ContentHandler in SAX when it comes to writing. Writing XML
streams via ContentHandler was very natural so they kept that half of
the model around.
> > Normally a navigator implies buffering, while the forward-only
> > cursor implies no buffering, but this really depends on
> > implementation of a particular XML store.
>
> Excellent.
My #1 favorite aspect of the .NET XML stack is the extensible design.
It's quite trivial to write custom implementations of XmlReader and
XPathNavigator over existing non-XML data stores or document formats.
This then makes it possible to use the built-in XPath, XSLT, XML Query
engines over such data. I've written some of these to expose the Windows
file system, registry, .NET assemblies, and zip files as Infosets[2].
[1] http://www.jcp.org/jsr/detail/173.jsp
[2] http://staff.develop.com/aarons/xmllinks.htm
-aaron
......................................................
. Aaron Skonnard, DevelopMentor (http://skonnard.com).
. Essential XML Quick Reference available online! .
. Download PDFs: http://www.develop.com/books .
......................................................
|
