Vista - Pros - Cons - Ideas/links on sending object graph through a web service using WCF.

We are pulling an object graph out of Nhibernate (it's a Object persistence
framework or OPF). The object graph consists of a basic object with about 5
properties for basic data types (ints, strings, etc), but also contains
about 7 properties that are Generic lists of other objects, and some of
those objects will contain a property that is a list of yet more objects up
to 4 or 5 levels deep. The object counts for each list will more than
likely be less than 100.

I have been a proponent of having a flat service/contract model (well a flat
model coming out of Nhibernate too, but that's another story) where the
service has atomic calls like the following:

Pseudo code

MyObject obj = GetObject(ID);
ChildObject1[] childObj1 = GetChileObject1(obj.ID);
ChildObjectLevel2[] childlevel2 = GetChildObjectLevel2(childObj1.ID).

The code could continue with calls up to 4 or 5 layers deep to grab parts of
the object graph depending on what you need. If the object graph were
needed, it could be built making many single calls on the service. Hope
this make sense.

One of the ideas here to transfer the whole graph at once by serializing all
the objects at once. I feel this is a bad idea due to the size of the XML,
and the client will need to be able to understand the XML (it's types
wouldn't be part of the WSDL import) and de-serialize it (using a SOAP
Formatter for example). This to me is like mixing "remoting" with web
services. What I do not know is whether the resulting XML from the SOAP
serializer is compatible with non-dot net clients. How much would a non-dot
net client need to know to re-construct the object graph?

Example WCF contract call:

String XML = GetObjectGraph(int ObjectID);

Thanks for reading this far..

To sum up, What's your opinion on transferring a complex object graph across
WCF? Yes (if yes, links examples?), No? Links/comments as to why or why
not.

Thank you!

Curt