My use-case is serializing compound DTO objects before sending them over the wire. There is no escaping the possibility that multiple references may exist. I can try to hack a solution that maintains references before/after serialization, but I really think this should be included in the serialization process as an option if at all possible.

There is already an older post on Stack Overflow about this, and I think (hope? ;) others would want the functionality as well.

http://stackoverflow.com/questions/11644905/serialize-deserialize-complex-type-to-json-with-object-references-and-dictionari)

I didn't want to bump that, so I'll add my comments here:

Currently, PCLs are significantly growing in popularity, as cross-platform development (Interoperability and mobility of objects) becomes increasingly important. Because of this trend, more complex object serialization because more important.

In my particular use-case, I have a transport object that contains some code inside of it to move among logical nodes (intra-process, inter-process, client-server, etc.). This is largely possible due to the shift toward PCLs, and I think I won't be alone in this kind of structure.

I would use BinaryFormatter if at all possible, but it isn't available in PCLs. Json.NET has problems with PCL implementations, and other serialization implementations I've found don't have the active development that Service Stack does - that's why I bought a license for the serializer in the first place. Active and good!

I know you have a ton of other features on the boil, but I think maintaining references in any of your serializers (not necessarily Json) would be a huge success in moving toward the PCL world. That's my three cents.