FW: Efficient XML - request for input

["Duane Nickull" <duane@xmlglobal.com>]

More insights.  There are some good comments here.

Duane

One insight: whatever might be true today could be completely
different tomorrow.

If the Java world is anything to go by, choosing your parser makes the
difference between how fast attributes vs. elements are parsed.  In
Perl, the Expat C library is used, and it's _fast_.

Probably regardless of whether you use attributes or elements, expect
a DOM tree to take up to 10 times the memory for the byte-size of your
document.  This number has actually been shrinking in the last few
releases (IIRC).

Since the software is going to be optimized far sooner than the
expected lifetime of ebXML, I would highly suggest that you focus on
using clear XML structures and not worry about immediate performance
issues.

To that end, it seems that more and more people are favoring
element-normalization (keep data out of attributes, use attributes
(sparingly) for meta-information).  Most new standards are also
favoring elements over attributes (XSLT, XSchema, etc.).

One small tip: if you have particular types of data that have an
inherent, simple representation as plain text, it's not necessary to
decompose it further into XML elements (that _will_ take up
unnecessary parsing and DOM space).  This can include comma-seperated
lists of symbols, email-addresses, dates, etc.

The SOAP spec <http://develop.com/soap/> includes a set of rules
(section 8) for converting application data types into XML that fits
very well with XSchema.  Keep an eye out over the next two months for
a clearer version of this spec.

  -- Ken

---
You are currently subscribed to perl-xml as: [duane@xmlglobal.com]
To unsubscribe, forward this message to
         leave-perl-xml-97975X@lyris.ActiveState.com
For non-automated Mailing List support, send email to  
         ListHelp@ActiveState.com