Re: [xml-dev] Optimizing XML performance

["Mukul Gandhi" <gandhi.mukul@gmail.com>]

By transmitting SAX events, I didn't mean sending the whole SAX trace
to the destination. That would be size inefficient, as you described.
I meant sending SAX events as they occur. By sending SAX events, I
mean sending the callback properties.

I think, using compression will work good over a slow network, but it
will need extra processing at source and destination (which shouldn't
be a problem, as these days processing speed is generally fast).

On 2/27/08, derek denny-brown <zuligag@gmail.com> wrote:
> I'm not sure how you expect serialized SAX events to even come close
> to the benefit you would get from compression.  XML tends to be highly
> compressible, so even ZIP or GZIP tends to get you a significant
> benefit at very low cost.  Serializing SAX events has the advantage
> that parsing is faster, but your concern was size.  Since SAX expands
> namespaces, a naive serialization of SAX events will often be
> significantly larger than the original XML.  All of this depends on
> what the capabilities of your WAN are.  Some studies we did at
> AgileDelta demonstrated that on modern hardware, 100Mb is fast enough
> that what you want is a fast parser.  At 10Mb you want to trade CPU
> cycles for compression, since a modern CPU would be waiting on netowrk
> I/O.


-- 
Regards,
Mukul Gandhi