RE: [xml-dev] Postel's Law Has No Exceptions

At 12:45 PM -0500 8/20/03, Hunsberger, Peter wrote:


>I don't think the situation works out that bad in reality: years ago I
>worked with a company that developed software for the exchange of X.12
>health care data and Blue Cross/Blue Shield ("standard") health care
>data.  Very regularly we would run into data produced by other vendors
>that did not conform to one standard or another.  I don't think we ever
>encountered a case where the vendor emitting the offending data was not
>willing to fix their software to emit conformant data...


The key in these sorts of examples is that you have to be able to 
prove easily and to everyone's satsisfaction that they are emitting 
nonconformant data. Draconian error handling (as well as the 
different issues of validity and schema-validity) in XML makes it 
very easy to prove this. When the problem is so obvious, then 
generally the offending party admiots there mistake, says "Sorry 
about that", and fixes it.

It's when you don't have draconian error handling and clear specs 
that people start arguing about whether the data is conformant or 
not, the problems don't get fixed, and you have to resign yourself to 
fixing the bad data for the other party.
-- 

   Elliotte Rusty Harold
   
   Processing XML with Java (Addison-Wesley, 2002)
   http://www.cafeconleche.org/books/xmljava
   http://www.amazon.com/exec/obidos/ISBN%3D0201771861/cafeaulaitA