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

E: Referee Comments (For Author and Editor)  
The paper describes a computational economy framework for resource allocation and for
regulating supply and demand in Grid computing environments. The paper is well written and
the ideas seem promising especially when it is applied to an environment like Internet. The
material of the paper has appeared in many other publications. It is very difficult for me
to see what is new in this draft.

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

This is an interesting forward looking paper which I would recommend
for publication after significant revision.
The economic model for computing resources is important and already
has had major impact on computer technology. For instance the trend to
using distributed computers (departmental machines) is partly due to
economic forces. Centralized Systems are difficult to manage
efficiently as if you charge then the machine is underutilized while
if any time is given away then it is saturated -- such tradeoffs
(and avoiding the tragedy of the commons with free time) is much easier
in small peer communities (such as departments)

I see the difference between time and cost optimized scheduling as
important and needing further research

That said, I found the paper disappointing. There was little discussion
of P2P concepts (trust, digital cash, reputation) but rather a
slightly superficial (IMHO) discussion of analogies with the real
world in section 3. Further I didn't really see much in sections 5 and
6 describing the technology and policy behind Grid economy based scheduling.
Note "trust" and "reputation" are critical -- they are basis of peer review
used in time allocation on most large computers.
Section 6 needs expansion -- one experiment is insufficient to make a point


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

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

This paper presents an economic model for evaluating which jobs run on
which resources. The paper is significantly flawed in several respects.

First, the topic. I think grid systems haven't achieved the maturity where
we can assign costs to resources and have users buy or trade them. We're
still grappling with technical and organisational issues in grids; economic
issues are distant. However, perhaps it is forward-thinking to talk about
economic issues already, so I don't view this concern as a major issue.

Second, the metaphors. There is a key difference between computing
resources and the stock market model that the authors present in several
sections. Computing resources are bought and sold so that they can be used;
stocks are bought and sold for future purchasing and selling. In that
respect, computing resources are more like groceries or electricity. It
doesn't make sense to talk about options, futures and trading.

Third, the scale. The authors seem to have ignored all issues pertaining to
the scale of transactions when there are several thousand hosts and several
hundred users. Bargaining or horse-trading between all these entities can
become a communications nightmare. Already a grid scheduler can become a
hotspot; with the additional burden of conducting auctions, it can miss a
high percentage of deadlines.

Fourth, the implementation. It's not clear that the section on Nimrod/G
actually contributes much to the authors' main thesis in the preceding
sections. The single experiment they conducted is hardly satisfactory
because there is no baseline comparison. From a scientific viewpoint, I
could not duplicate that experiment if I had their resources because they
haven't given me all the information about the experiment.


F: Presentation Changes

The paper is verbose, extremely so. Presenting definitions of well-known
terms like "economics", "monopoly" and "oligopoly" is patronising to
readers. Repeating the ideas of resources being offered and purchased is
annoying. There are spelling mistakes ("Cutomised") and grammatical errors
(unbound referrents). Citations are treated as first-class language
entities, which they are not.

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