Hi Rick,

-------------
[RS] I'm not sure if I buy that... the point isn't so much that you need a DOM, but that in order to provide XSLT and XPATH you need to have the whole document in most cases in order to do a realistic selection of nodes.If you do SelectNodes() on a nodeset that runs to the end of the doc with subs you're essentially walking the entire document...

But if more complex searching or manipulation is involved I doubt that there's much benefit to doing it.

-----------------

I don't know what you mean by "realistic" here. A selectNodes() on a large XML document is *not* the right way to do sophisticated searching or querying on it. Indexes in XSLT do work <s>.

Maybe I'll have a chance to change your mind about this sometime!

However, IAC:

---------
[RS] Finally we know that using huge XML documents should be avoided as much as possible and atomic data be used wherever possible.
------------

Nobody would ever dispute this, Rick, least of all me <g>... and:

--------------
[RS] With moderate data size DOM performance is more than adequate and doesn't compare even badly with something like an XMLReader.
--------------

That's true also. I'm not even saying that SAX is superior to DOM, if you like DOM and the performance is adequate there is no reason for you, me, or the people who started this thread not to use it <s>!

But the reason I came on the thread was because it isn't correct to say that XML *itself* is in some way restricted to small datasets. Certain methodologies are *presently* limited, yes, but XML is not inherently limited.

16 Mb are a *reasonable* size XML document to handle in some scenarios. It's true that atomic exchange of data is preferable where possible, but there are lots of data-consumption scenarios where it's just not possible, and XML still makes sense because of unlike data systems. Not all the time, but sometimes it does. And it's quite possible to do it.

>L<