On the Design of a Practical Information-Centric Transport

AbstractA recent chain of exploratory publications proposed information-centric architecturesfor the Internet. The common pitfall of such proposals is the imbalance of upfront costsand immediate bene ts. To address this concern, we focus on prospects of piecemeal adop-tion. We start with the necessary basic primitives of any infocentric architecture, primarilyself-certifying names, to determine what are they sucient for. We de ne natural inter-faces to other parts of the architecture, primarily naming and routing, for which no specialassumptions are made. We de ne a natural separation of infocentric transport and internet-working layers and their message vocabulary, that allows to run our infocentric transportover IP, UDP, TCP, HTTP or entirely IP-free. As a proof of concept, we have implementeda UDP-based transport protocol named swift with per-datagram data integrity checks. Ourarchitecture highly prioritizes modularity and sketches a path for piecemeal adoption whichwe consider a critical enabler of any progress in the eld.Victor Grishchenko, Johan Pouwelse and Henk Sips are employed by Delft Universityof Technology. Flutra Osmani and Raul Jimenez are employed by the Royal Institute ofTechnology, Isafjordsgatan 39, Kista, Sweden, fflutrao, rauljcg@kth.se.