>>> If someone builds a databinding tool I don't care if they implement 
the 
>>> key/keyref system.
>>
>> Why is key/keyref any less important for a databinding tool?  If a 
>> key/keyref constraint is required for an instance to be meaningful, 
then 
>> why would I want a databinding tool to marshall/unmarshall an instance 
if 
>> it didn't satisfy that constraint?
> 
> Any number of scenarios-- the most likely being that a validation 
> pipeline is in front of the databinding/serialization tool. There are 
> many cases where I don't need or want my data-binding tool verifying the 

> integrity of the data passed to it-- the duplicate checks could be 
> time-consuming. If the tool is designed to consume PSVI than it's 
> early-binding support of features like key/keyref would be pointless.

Given that response, shouldn't you have originally said something like:

        If someone builds a databinding tool I don't care if
        they implement any form of schema validation

rather than signaling out key/keyref?

When I have heard other people make statements like your original they
have always meant "don't need to do key/keyref but must do simple type,
UPA, etc. validation" and that is how I interpreted your original
statement...which is why I asked why key/keyref validation was any less
important than simple type, UPA, etc. validation.

pvb

From mike@s... Thu Aug 04 21:51:34 2005
Received: from mag