********************** Referee Number 1 **********************

The paper provides a qualitative and quantitative comparison of a
concrete CORBA ORB (Borland VisiBroker for C++) and Message Queue
System (IBM MQSeries for Windows 2000). This comparison is of
interest and of obvious practical relevance to the community.
However, the test carried out are only simple client-server (ping-
pong) test and cannot reflect the bahaviour of more complex programs
(e.g. of advanced patterns like call-back and so on). Moreover, the
generalisation of the results to CORBA ORBs and Message Queue
Systems, in general, is doubtful. The paper is a typical short paper
for me.

Section 1 introduces to middleware in general and to the project
carried our by the author in special. It is well written.

Section 2 reflects some key-dates of networks.

Section 3 introduces to CORBA and message queues. This Section can be
shortened, as it is not more than a basic introduction which is given
in Distributed Systems Course.

Section 4 describes the experimental settings.

Section 5 presents detailed results. The major contributions are,
however, not suprising: to put it ironically, you say (obviously much
more elegant) with higher packet loss performance degrades and CORBA
is better then. The problem here is that this Section must attract
the reader more. Use other programs, more complex, which uses
distributed patterns (factory, call-back, etc.), how can they be
realized with CORBA, message queues, what does it mean to performance
under the various conditions.

Section 6 discusses some general issues again. I was a little bit
confused about the placement of this Section.

Section 7 presents the conclusion. Here, the author presents us with
some information on QoS requirements for speech and video delivery
and argues that systems have to better fulfill these requirements.
However, this confused me in this context. Is the focus of this
paper: real-time applications. If so, why do you compare CORBA and
Message Queues. Instead you have to compare real-time implementations
of CORBA ORBs, you should test it on different network technologies
than ATM. Or is this paper general-purpose application oriented as I
understood from the introduction.

The last sentence is rather frustrating to the reader : "The
selection of different middleware technologies cannot considerably
improve the response time for remote object invocations.". So the
comparison was meaningless, and only more reliable networks can solve
the problem ?

I think, turning again to the real-time problem that there is
considerable difference in middleware, for instance serialization
time differs considerably in the middleware used and this has been
examined elsewhere.

Finally, the list of references is very small with respect to the
problem dicussed. There exists a lot of comparison of middlewares
(quantitatively and qualitatively),  see previous issues CandC:PandE.
How is your comparison novel to these ?


F: Presentation Changes 
Several  major items have, to my opinion, to be updated before
publishing :

1.	more experiments : use of more complex programs using various
patterns.
2.	shorten the systems descriptions where they are basic and be more
detailed for the comparisons.
3.	relate to other works done (epsecially in the comparison of
different middlewares).
4.	clear the application context (real-time ?)