Referee 1 *******************************************

E: Referee Comments (For Author and Editor)
Technically this was an excellent paper with a very thorough treatment of
the issues pertinent to Grid event services. The prototype system described,
while significantly smaller than the billion node system mentioned in the
introduction, was complete enough (in both hardware and software) to
exercise the methods and algorithms outlined in the paper. Whether the
architecture can scale in size and complexity (dynamic brokers, security,
etc.) remains to be seen but this paper represents an excellent beginning.
One technical point that I am curious about, is whether other routing
criteria besides path length have been considered by the authors (e.g.
aggregate link cost, speed).

My only criticsm regarding this paper is that I think that it needs to be
edited by a technical editor before it can be considered ready for
publication.

Referee 2 *******************************************

E: Referee Comments (For Author and Editor) 
------------------------------

  The idea an the concept of a distributed event service and
the distributed broker topology is good and I have no
criticism on the technical content of this paper. I think 
the solutions fits very well to actual structure of the Internet
with its topology of nets and subnets, using also routing
for message/package delivery and its capabillity to handle
non-working parts of the network.


  For the special issue on Grid Computing Environments,
I recommend, that you present only a shorter overview of GES
and that you add some remarks on how do you think it may fit
into existing Grid infrastrutures and computing environments.

  In the paper, you mentioned the prototype implemenation.
Is this already used by other groups, or is it available
somehow? This information should be also included.

Referee 3 *******************************************

Overview:

This paper presents a hierarchical architecture for event distribution in the
upcoming peer-to-peer and Grid-oriented networks. The paper describes requirements
for such a system and the underlying components and mechanisms. The overall
architecture is based on brokering. The event delivery solution is a mixture of Java
Messaging Service and Quenching techniques. Problems related to the reliability of
event delivery is addressed by allowing broker nodes to be lightweight and to
permanently fail, and introducing stable storage nodes that need to recover from a
failure in a finite amount of time. Different protocols associated to the
architecture, i.e. Gateway propagation protocol, Event Routing Protocol, are
presented.

Organization:

The overall organization of the paper is very good and it covers the topic quite
well. The first half of the paper (up to the description of profile graphs) reads
well. The second half however is hard to read. The extensive descriptions given for
each requirement tend to take the focus away from the problem addressed by the
paper. A clear statement of what the authors consider the contributions of the paper
are will also help.

Technical Content:

In this paper, the authors choose a hierarchical architecture based on a
publish/subscribe model in which broker nodes in the network are used as intelligent
routers, and storage nodes handle each client's storage needs for
synchronous/asynchronous event passing. Alternatives to this model such as the
Scribe (http://www.research.microsoft.com/~antr/pastry/) approach or the JXTA (http:
//www.jxta.org) approach are not really described in the paper (only quickly
mentioned as related work), leaving the reader with a lack of pointers that could
help placing this work with respect to others in the area of event/messaging
systems. Most of the technical discussions are in depth and supported by examples
having figures as references that are helpful for visual understandings of the
issues described. Some explanations however do remain unclear and seem to be dropped
there just for the "look" (e.g. Clustering coefficient -2.3; gateway propagation
protocol -3.1.1; profile-graph). Also, as mentioned earlier, some of the
descriptions tend to add excessive details that take the focus away from the problem
being addressed.

Innovation/Impact:

Despite their key role in every network systems and particularly peer-to-peer
oriented architectures, event dissemination mechanisms don't seem to have received
much attention in the literature. This paper does a good job addressing this gap and
provides a good reference point for the topic.

References:

The bibliography is extensive - however most of the citation could use more
discussion as to why they are relevant.

Recommendations & Presentation changes: 

From an organizational point of view, it would be nice to have some of the details
moved into an appendix or to a separate technical paper, allowing this paper to be
more focused. Also, the links in figure 1 are hard to differentiate and only figure
1 (out of 3) includes this legend. Finally, I found the profile graph section
particularly hard to understand - it could do with some reorganization.