This list summarizes the discussion open points collected from multiple sources including feedback from meetings, off-line interactions with individuals, various exchanges with experts. We list them in order to obtain a wider feedback as well as different viewpoints. Feel free to share your arguments with the LCCN community.
Discussion Open Points
|Potential misuse of the learning capability||Which measures to take and how to avoid that individuals can potentially misuse the learning capability|
|Positioning wrt Management Plane/Management System||Learning-based control focuses on augmenting current control process(es) underlying e.g. routing protocol modules (OSPF, BGP, etc.), traffic control, etc. However, the (open) question remains that "augmenting the control system" could overlap with specific operational processes currently driven by the management plane/system|
|Potential impact of learning capability introduction compared to the current control processes||What are the limits in the delegation of control to the network systems/devices themselves ? How to handle the relative complexity resulting from the introduction of learning capability on network systems (vs intended functional and performance gain) ?|
|Does the end-to-end principle prevents increasing "intelligence" on intermediate systems ?||This additional control capability does not aim to interfere with the end-to-end traffic exchanges: The exception is when these are becoming detrimental to the majority. In other terms, the question can be stated as what / who determines when these effects are potentially detrimental. This an open question for which the answer depends on application. For instance, for network self-defense systems the answer is different from congestion control and from the routing perspective.|
Showing 4 items