In article <> you write:
>We have all seen the various styles of WXS schema in use: Russian
>Doll, with everything nested, Venetian Blind, full of global complex
>type definitions, and Salami Slice, with global element definitions
>accessed using "ref" attributes.
>
>Has anyone considered the effect of these different models on parser
>performance, either from a theoretical or practical view?
I wouldn't expect any noticeable difference. In XSV we generate a
finite-state machine for content models, with the states having
pointers to the element declarations and the element declarations
having pointers to the type declarations, so there is no looking-up of
element or type names in the usual case (there is for wildcards of
course).
A lot will depend on whether you end up reading in the schema for each
document. That may well take much longer than the validation itself
for small documents.
-- Richard
