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

E. Referred Comments to Author and Editor

This paper presents scenarios that motivate research in developing new
models for multidisciplinary science.  Multidisciplinary communities are
characterized and a high level architecture presented.  Five existing
problem solving environments are described. Several core
technologies/approaches for building multidisciplinary GCEs are then described.

This paper contains significant interesting and applicable content
especially sections 1.3, 1.4, and section 3.  It could be improved by
condensing it significantly, especially section 2 which is not well
motivated (I'm going to tell you about 5 PSEs because...?) and with the
exception of S4W, not tied in well with following sections.  A reader
will be tempted to skip it.  The reader will not become experts in
these areas or applications.  The depth of discussion distracts from
the informative content of the paper.  Since the paper is quite long, I would
suggest discussing only one in depth and capture the key points about
the others in a short paragraph.

References need to be ordered.

The abstract could be "abstracted-up" one level.  It currently reads more
like an outline or roadmap than expected.

Grid computing is a broad term and should be defined for the context of
the paper.  The first sentence of the abstract asserts the maturity of
grid computing.  This is not an easily accepted statement.

In the opening of the introduction, a shift from lower level application
scheduling and execution to higher level problem solving is asserted.
Its very debatable whether this is happening yet.  The reference to "The
Grid" book makes it even less convincing since it was published in 1998 and
the early work to develop infrastructure was just starting. Also, the
paper is really about early research into supporting higher level problem
solving so it seems to contradict this statement.

Good motivating content in 1.1.

Sections 1.3 and 1.4 reflect substantial work and derived concepts. As
such aren't appropriate content for an Introduction section.
The content is good however.

Section 3.1 last sentence - this is a strong statement without any
references.  It's not clear what the "traditional approaches" are.

Section 3.1, third to last paragraph.  "The modeling and performance
requirements in a multidisciplinary GCE mean that both approaches are too
restrictive." Again a strong statement.  Is this proved?  Also the reader
has to go back and search for what two approaches are being referred to.
Perhaps this is because the paragraph contains too many references for
the first category.  I would suggest that the approach used for the
second category (i.e. just reference numbers) would be sufficient and
allow the main point to come through.  

Figure 12 and the couple sentences referring to it could be left out.  At
this point in the paper, domain specific content confuses the discussion.


F. Presentation Changes

The style seems a little informal for a journal - for example the
repeated use of "we" in 1.1 paragraph 1 and "we" and "our" in section 1
in general.  This portrays casual conversation about experience.

In general, it would help to talk to the figures more. The reader has to
do quite a bit of work to figure them out and make the connections to
text.  Make it easy for the reader.

Forward and backward references (e.g. See section xxx) didn't help and 
were distracting. I suggest they be removed.  Organize content or
rephrase so they are not necessary.

The attempt to cover too much ground and refer to too much other work
lessens the paper by reducing readability and drowning the main points.

Last two paragraphs of the discussion are hard to read and don't add to the
paper.  Are discussions of future work appropriate for a journal article?
I suggest they be removed.


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

Referee Comments 

The paper is well organized and presented. The goals of the paper are clearly outlined and
motivated using scenarios - for e.g. in scenario 2 they mention that the various components
in GCE can be geographically distributed thus giving importance on collaboration, while the
importance of compositional modeling in scenario 3. 5 examples of PSE are presented.

The system support section is technically sound - the first subsection brings out the
importance of representation and it clear presents the their representation based on BSML.
However, the format conversion described however is not clear. In the first sentence of the
section "Format Conversion and Chain Management" the authors write "one of the benefits of
semi structured data representations is automatic format conversion" but they do not
elaborate more on it.

Sieve and Symphony are the most interesting contributions of the paper - however the paper
only presents an overall picture leaving unanswered questions. For e.g. how does Sieve know
about the data sources? How do the users enter the collaboration mode? In Symphony the
present experiments are done "based on a specially created server" - no more details are
provided on these experiments. How does the user know where the failure occurred in a
simulation - there seem to be no concept of logs for a simulation? A little more detail in
these subsections will help reader understanding the PSE framework better.

Presentation Changes
Figure 13 is not very clear. A sharper image will definitely make it clearer and easier for
the reader to understand. There are some typos - for e.g. Section 2.2, second paragraph has
"planform" instead of platform.


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

E: Referee Comments (For Author and Editor)

The paper provides a clear and interesting insight into the issue of 
building grid environments for multidisciplinary communities, providing 
typical scenarios and common problems, followed by the description of 
several problem solving environments built by the authors and the more
elaborate discussion of their contribution.

The only serious concern that I have is the length of the paper, which
is currently 32 pages or, excluding references and the table of
contents (sic!), 23 full pages of text and figures.  And the paper
doesn't currently use the right style for typesetting.  Once it's
converted it will probably be more like 40 pages.  While there were 
no page limits in the call for papers, I do think this is a bit much.
All sections of the paper could be shortened, partly by omitting 
examples, partly by being more concise.

While the paper contains an impressive number of references to the
work of others, a separate "State of the art" type of section briefly
describing other similar efforts would certainly be a plus.

This is actually not very relevant, but section 1.3 (GCEs for
Multidisciplinary Grid Communities: Characteristics) mentions "gcc
2.0.8".  There was never such a version.  See
http://gcc.gnu.org/releases.html

F: Presentation Changes

As already mentioned, the paper doesn't currently use the right style
for typesetting.

The quality of most of the figures generated from bitmaps (3, 4, 5,
13) is definitely unsatisfactory: they are so blurred that it's hardly
possible to see anything.

I spotted a single typo: "deployment of of JavaBeans"
(section 3.6).