Recursive (or mutually recursive) types are sometimes desirable. Consider an organization chart, or the descenants or ancestors of a person in a genealogy application. There is no reason why the serialization code shoould crash with a stack overflow since the data actually sent clearly has a recursion which terminates.
Cycles using IDREFs should be handled using the multiref construct in serialization. It seems to me a bug if this fails.
Jeff
Anne Thomas Manes wrote:
Don't put cyclic loops in your schema.
On Apr 12, 2005 4:39 PM, Soti, Dheeraj <[EMAIL PROTECTED]> wrote:
Hi,
I have a schema which has cyclic loops and AXIS goes into a circular loop while trying to serialize the objects (throws Stack overflow exception). I can write my own serializer/deserializer but I am confused that how it is going to work on client's end. In addition to WSDL wouldn't the client also need the serializer/deserializer? Till now I have been thinking that the best approach is to just write a WSDL and that is the only thing any client would need to get access to my service. If I make sure that it is interoperable then they can generate client side proxies in any environment and they should be ready to go. I can generate server side classes using AXIS or anything other tool.
Am I correct or I am missing some very basic concept here?
Thanks
Dheeraj
