We have not yet designed the negotiation protocol because we need more experience with applications that need it. In particular, it is not clear whether the proxy should provide support for real-time constraints, and to what degree. Full-fledged real-time support complicates the design of the proxy. It is possible that for most Active cache applications that need latency guarantees (for example, distributed collaboration and multimedia streaming), the proxy needs only guarantee prompt execution of the cache applet upon client requests and periodic execution of the applet (via an alarm service). We plan to evaluate various design choices as we develop those applications.
Currently, resource management policies in Active Cache are still research topics. In our current prototype, all documents are un-negotiated objects. We use the simple LRU policy in deciding which document to cache, regardless of whether the document has associated cache applet or not. We also impose fixed constants on the resource consumption of all cache applets. These choices are purely for the ease of implementation, and we plan to refine them soon.