[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: raw storage about NXD

• To: xml-dev <xml-dev@l...>
• Subject: Re: raw storage about NXD
• From: Michael Champion <mc@x...>
• Date: Tue, 23 Dec 2003 10:30:24 -0500

On Dec 23, 2003, at 7:48 AM, Qiang Song wrote:
>  I am
> still not very clear about the raw storage form for XML
> documents.is it in pure text format?If true,how to execute the query 
> and
> update operations.how to index for documents.if using DOM to build a 
> tree
> from pure-text XML document before executing query or update,then i 
> think
> it will be very low-efficiency.
>

I'd just echo what Michael Kay said -- doing this well is hard, and the 
companies who work in this area generally do not say much about their 
internal technologies.  The point of "native XML" databases is NOT that 
the data is stored internally in a standard XML text or Infoset/DOM 
format, but that one uses XML interfaces such as DTDs/Schemas, 
XPath/XQuery, and SAX/DOM to interact with the DB *as if* it were just 
a huge text file or DOM tree.

For details on how people have addressed these challenges, you should 
probably look at the open source NXDs such as Apache Xindice, dbXML, 
and Sleepycat.

Ultimately, each implementation makes engineering tradeoffs that make 
their approach more or less optimal for small vs large quantities of 
data, large documents or small documents, XML representing unstructured 
documents or structured data, optimized for reading or optimized for 
updating, etc.  The only real way to determine whether a specific 
product (and the technology it is built on) is suitable for a specific 
application is to prototype it and run benchmarks with realistic data 
in the actual hardware/software environment.  Don't believe ANY 
marketing or benchmarking about performance of NXDs vs one another or 
vs alternatives until you verify it under the conditions that YOU 
expect.

• Prev by Date: RE: raw storage about NXD
• Next by Date: Re: patterns and restrictions
• Previous by thread: RE: raw storage about NXD
• Index(es):
  • Date
  • Thread