This paper points out that middle-boxes and Internet architects have never been happy with each other and given the fact that middle-boxes have become enormously useful, integrating them into the Internet architecture is an important thing. To that end, they propose Delegation Oriented Architecture (DOA).
DOA achieves its goal using flat namespace of self-certifying names and the ability for the sender and receiver to define the intermediaries that process the packets. Routing between end-point identifiers (EIDs) using a global resolution service.
The reason for doing this seems unconvincing. Architectural cleanliness seems a good thing but given that NATs and firewalls have been deployed and work fairly well, it seems unnecessary to deploy this new addressing format etc. in DOA. At least, I am not convinced why someone would go for this. Peer-to-peer applications have also fixed this problem through an externally accessible machine. Even if you were to disregard all this, the whole system has serious performance concerns dependent on the DHT resolution.
Overall this paper has a highly "academic" feel to it and I wouldn't vote to keep it on the list.
Subscribe to:
Post Comments (Atom)
3 comments:
Interesting perspectives. In general, I also prefer the i3 work as being more general and focusing on proposing one mechanism that can solve several problems at once. However, the implementation challenges are considerable, as Ion admits in the paper, and his approaches are feasible even if they are left for the future. Unclear where i3 is going, though the ideas were used in Coral I believe.
Are you getting a little behind in your blogging?
Yes, last week was bad...TA work, TA work! :-) I hope to catch up this week...
Post a Comment