As anyone who has worked with converting XML schema into usable class libraries knows, sometimes the resulting proxy classes don't capture all of the details in the schema.
For example, an attribute might be defined in the schema as an integer, but might be converted to a string in the class.
Or an element might have a min/max-Occurs, or an attribute could be defined as optional, with both of those definitions being lost entirely in the class conversion. Other XML constructs, such as sequences and choices get flattened into less meaningful arrays.
How do people deal with this lack of adequate conversion? Do you add back in proper validation code? Ignore it and let the remote server throw back errors to iteratively correct?