Paper Review:
Integrated Services in the Internet Architecture: an Overview
Reviewer: Robert Dugas
Problem
The problem addressed in this paper is that of adapting the current IP layer to support various different service types such as real-time
multicast and others in addition to the point-to-point best effort service currently
available.
Contribution
The foundation for such an infrastructure is laid down, both by citing specifying the interface for
such a differentiated QOS mechanism and by explaining how such an interface might be realized.
Main Ideas
- Suggests that in order to support real-time interactive applications,the IP infrastructure must be extended
- Develops core functionality necessary for such as system
- Packet scheduler
- Classifier
- Admission control
- Reservation setup protocol
- Cites examples of implemenations of such mechanisms as options for the realization of such an extended network infrastructure
Critique
Significance:2
None of the technical ideas set forth in this paper had not been mentioned elsewhere before. Thus, the significance
of this paper derives primarily from the notion that, okay we've got all these ideas for extending the
IP layer, here's why we need to do it, here is the form it could take, let's get going.
Methodology:
There really is no methodoly in this paper. In fact, it is continually self-referenced as a "memo"
rather than any type of research paper. It is mainly an organization of previous ideas from a
variety of sources and a call for action on the part of the internet community.
Limitation:
Inherent in the utter lack of real-world methodology is the limitation that we have no
idea how the proposed system may or may not actually function. Further, at one point
the authors concede that the proposed archeticture would introduce pricing and authentication
issues on an unprecedented scale. They simply gloss over this by noting that it is an issue
which must be "examined futher."
Lessons:
It seems like the major lesson of this paper is that end-to-end systems are inadequate for
the demands of real-time applications and thus we need to extend the ip layer to provide such differentiated QOS support.
However, the validity of this lesson is dubious as, since 1984, numerous applications have been developed which
do just that, albeit with varying degrees of success.