DNS Paper Review

The Paper reports on the development life-cycle of domain name system from its ancestor system in DARPA network. The authors address the cause of switching to such a distributed system and analyze its successes and shortcomings in its concrete deplyment.

The main strength of the paper is that it provides a comrehensive set of reasons for designing the new DNS features; e.g. the scalability issue of the HOSTS.TXT name system had a driving role for switching into a distributed hierarchical cache-enabled NS. The historical life-cycle of the development is well described to the extent that is seems the authors were in the main stream of DNS early-phase developments. A major concern of DNS designers is well reflected in the paper: If you distribute the control of the Naming system, you'll lose much expertise and responsibility, unless you have strict distributed control mechanisms. The tradeoff here is that if you go towards extreme control mechanisms, you'll in fact tolerate the many issues of a centralized system, like extreme maintenace cost, ... . One approach for resolving this issue might be to create hgih competition among organizations providings domain name system resolvers by increasing the fee of resolving in the Internet community. Altough this may decrease the Internet's speed of growth, however, by un-radically using it, it might lead into a higher quality network than what we have.

One major drawback in the work of DNS designers, at least as far as this paper mentions, is the lack of a prototype implementation, whether experimentation or simulation, before deploying the DNS system. Many of the *surprises* might be discovered in such a prototype before incuring a huge cost after deployment. There are also some issues that the authors do not address like "Security" of DNS which seems to be from the very first concerns in a distributed system in comparison to its cetralized HOSTS.TXT father. Minor issues like handling the heterogeneity of DNS resolvers e.g. different bandwidth and their impact on the system (like congestion) is not addressed well too (Although the system's design seems to be robust to this specific issue).

The paper's presentation is not that well; it could be much simpler, yet communicating the whole concept.
Received on Mon Oct 03 2005 - 10:49:55 EDT