Don talks about The dark side of tolerance. While there is no standard that I know of, I have been using a homegrown header for warning users of bad XML. I’ve built some fairly tolerant web services in the past few months for my own research.

For example, I allow you to post data as an element or an attribute, and I’ll grab the data from which ever one is present. If both are present, I grab the attribute and warn them that there was duplicate data in an element.

If they pass in elements or attributes that I don’t understand, I list them all out in a return header for them to see.

This all works pretty well. Right now I’m only handling extra data, duplicate data, and pulling data from multiple sources (element or attribute).

I did all this to answer a question. How tolerant should I be? [News from the Forest]

I’m still not convinced.  This related to the loosely coupled web services discussion we’ve been having, and I’m still not convinced that we shouldn’t be doing a schema validation on the input to our web services.  If you want extensibility, use a xsd:any extensibility node; if you want parameters to be optional, mark them as such with minOccurs=’0′. 

Justin, what is your reasoning behind wanting to have your web service be tolerant of what might be described as malformed input?