RE: [xml-dev] Article on nytimes.com about Microsoft

["Alessandro Triglia" <sandro@mclink.it>]

> -----Original Message-----
> From: Alexander Philippou [mailto:alex@noemax.com] 
> Sent: Thursday, September 06, 2007 05:04
> To: 'Alessandro Triglia'
> Cc: 'Michael Champion'; '[Public XML Dev]'
> Subject: RE: [xml-dev] Article on nytimes.com about Microsoft
> 
> Hi Alessandro,
> 
> All measurements were made without using external 
> vocabularies, encoding algorithms or restricted alphabets.
> 
> For compactness comparisons, the freely available FI 
> Converter and MSBXML Converter utilities were used in their 
> default settings. For performance benchmarks, Microsoft's XML 
> Mark benchmark kit was used with only the minimal 
> modifications required in order to test each XML encoding. No 
> special tweaking or performance boosting tricks anywhere.


Thank you for the information.

That implies that, in a real scenario, it should be possible to achieve an
even better performance for some of those XML instances by using external
vocabularies and/or encoding algorithms.

For example, I recall that the X3D standard mentioned earlier uses an
external vocabulary specified in the X3D standard itself.  For short
documents with a somewhat predictable markup and content this avoids the
transmission of a certain number of bytes on each instance, and allows the
use of pre-initialized string tables both in writers and in readers, thus
avoiding some waste of time creating those tables dynamically on both sides.

The use of encoding algorithms may be very effective for documents
containing a large proportion of character sequences that match certain
common patterns.  Unlike external vocabularies, the use of standard encoding
algorithms does not rely on any information external to the document.

Alessandro


> 
> Alexander
> 
> > -----Original Message-----
> > From: Alessandro Triglia [mailto:sandro@mclink.it]
> > Sent: Thursday 6 September 2007 03:05
> > To: 'Alexander Philippou'
> > Cc: 'Michael Champion'; '[Public XML Dev]'
> > Subject: RE: [xml-dev] Article on nytimes.com about Microsoft
> > 
> > 
> > Hi Alexander,
> > 
> > Those performance numbers are impressive.  Did you use any of the 
> > special features of Fast Infoset such as external vocabularies or 
> > encoding algorithms in your tests in order to maximize 
> compactness and 
> > speed for some documents?
> > 
> > Alessandro Triglia