Is today's DNS the right solution for middleboxes selection?

In-network services, often implemented using middleboxes, are key components of today's network applications (e.g., CDN, antiviruses, proxies, etc.). Current solutions for the selection of a serving middebox assume the presence of a single middlebox on the end-to-end network path, leading to not optimal solutions in terms of network performance when multiple middleboxes are inserted i.e., multiple parties are involved. This paper highlights the cost of a non-optimized middlebox selection strategy and suggests directions for further investigation.