[Rets-dev] 2.0 overhead redux

[dbt]

On Wed, Apr 25, 2007 at 02:24:42PM -0600, Eron Wright wrote:
> I was not involved in the original discussions, but a few things that come to mind are:
> 
> 1.	
> 	Attributes have different whitespacing and encoding rules.  For large bodies of text, elements are more efficient and do a better job at preserving whitespace.  For small values and codes, attributes make good sense.

Indeed.

Fun bit of XML trivia: It's illegal to represent a <, escaped or
otherwise, inside an XML attribute value.  While most XML toolkits
don't barf on this, there's no reason for a conforming xml application
to accept it.

This makes attributes useless for arbitrary user-provided data.

to address an earlier point, plenty of json applications represent
tabular data with:
{data: [['col1', 'col2', 'col3'],
    ['val1a', 'val2a', 'val3a'],
    ['val1b', 'val2b', 'val3b'], etc 
] }

This would eliminate the duplication of keys.