Subject: CCPE Portal C558 From: "Stephen Mock" Date: Sun, 23 Dec 2001 11:32:00 -0800 To: gcf@indiana.edu, mthomas@sdsc.edu, "Kurt Mueller" , "Maytal Dahan" Geoffrey, Here is the revised version of the Perl CoG paper in pdf and MSWord format. I have attached the figures which are diagrams or pictures as separate files as well. (there are several figures which are code examples which I've left in the body). I have also attached a text file on how the referees comments were addressed. Please let me know if I missed anything. Thanks! Stephen Mock Group Lead, Grid Portal Architectures Grid and Cluster Computing San Diego Supercomputer Center > -----Original Message----- > From: Geoffrey Fox [mailto:gcf@indiana.edu] > Sent: Saturday, October 20, 2001 2:38 PM > To: Mary Thomas (C557 C558) > Subject: Referee Reports on GCE Special Issue of CCPE: C558 > > > I enclose 2 referee reports. I think you can address these > straightforwardly > by adding more detail > I hope that you will submit a revised version. > > I would like to receive your revised paper by the end > of November. Please return to me by EMAIL > a) PDF version of revised paper > b) Text Word or PDF comment on how referees' comments addressed > c) In spite of what journal says, I do not need at this stage > copyright release. You will send this directly to Wiley in > England after they contact you. > > Please put in subject of EMAIL > CCPE Portal Cabc > where Cabc number of your paper > Thank you for your contribution to this excellent special issue > > Referee 1 ******************************************* > > E: Referee Comments (For Author and Editor) > ------------------------------ > This paper is a very interesting short sketch of a developing > project. It describes quite well the modules and interactions with > other tools. What is missing is reference to related work which would > help to see what is so specific about the project. Furthermore the > concepts and ideas behind the modules and tools have to be added. > The authors mention their object oriented approach at several places, but > they don't elaborate it further. > > If this short sketch is worked out into a full paper it may be > very interesting. > > > F: Presentation Changes > > The paper is not formatted according to the guidelines of CPE special > issue, an abstract is essential. > The authors should read the instructions at > http://www.interscience.wiley.com/jpages/1040-3108/ and get the > appropriate style files. > > More information about the similarities and differences between the > Perl CoG and the other CoGs should be provided. Do they have the same > design and architecture? What are the reasons for possible > differences? What is special for Perl? > > Figure 1 shows the architecture of a portal developed with a different > toolkit (GridPort). The authors state that the architecture > would be the same if the Perl CoG had been used. Nevertheless the > relevance of this figure for a paper about the Perl CoG is unclear. > Would the architecture differ if a non Perl toolkit had been used for > the implementation? > > Since the Globus client utilities are required for the Perl CoG the > statement "the requirements for the Perl CoG are minimal, and the > installation is simple" is a contradiction to the earlier statement > "the installation of the Globus client utilities may present some > challenges". > > The whole section 3 resembles more an installation instruction than a > scientific paper. We recommend to remove this section and to include > the relevant information provided here in the description of the > architecture. > > In section 4 the handling of interactive and batch jobs are > described. In the presented context is is not clear what the > differences between these two kind of jobs are and why only the batch > jobs have been wrapped in a object oriented module. > > The authors should describe what distinguishes the RSL module from a > string class library. What abstraction level is provided to the user? > Is it the responsibility of the user to get a correct RSL string or is > this guaranteed by the RSL module in an OO style fashion? > > Only the search and RSl modules are pure Perl implementations, all > others are simple wrappers to the underlying Globus utilities. Since > the potential benefit of a pure Perl implementation is huge and also > mentioned by the authors it would be very interesting why this > approach was taken by the Perl CoG developers. Was it a matter of > convenience for the developers or are there more fundamental issues > involved? > > Referee 2 ******************************************* > > Overview : > > This paper describes a Perl implementation of a Commodity > Grid Toolkit ( Cog Kit ). The paper explains why such kits > are needed and why a Perl implementation is necessary. > > Review : > > Having seen presentations on this excellent project, this > paper does not seem to do it justice. Basically the paper is too > short to adequately explain the various components and the design > decisions. Perl is probably the most widely used scripting language > and is absolutely appropriate for system level tasks. The marriage to > Grid technologies is natural. Some of the advanced features of Perl5 > could be explained better ( where appropriate ) that show why it's > a stable yet modern scripting language. A more in-depth explanation > of the Perl CoG Kit models is necessary. > > Module creation is discussed. A utility to convert 'C' headers to > stubs is currently being used. As mentioned in the future plans, > a move to SWIG is planned. This is a wise decision. > > Basically, this is an excellent project and a Perl implementation > is extremely attractive. The paper just needs to go into a bit more > depth on the details. > > Recommendation : > > Publish with modifications. > > -- > : > : Geoffrey Fox gcf@indiana.edu FAX 8128567972 > : Phones Cell 315-254-6387 Home 8123239196 Lab 8128567977 CS 8128553788 > > Subject: Referee Reports The main comment from the referee reports was the lack of detail about the architeture of the Perl CoG kit. I flushed out the details on the architecture with respect to the grid, the module architecture, and the concepts, ideas ,and choices made for for the individual modules. This includes going into more detail on why some modules were developed initially as wrappers to globus utilities, and why some are Pure Perl. I also added simple code examples for most of the modules. I described in detail the difference between the two job modules found in the Perl CoG. There was some lack of clarity on what the difference was between the interactive and batch job modules. I ironed out a possible point of inconsistency about the requirements of the CoG. I attempted to disperse the 'requirements' section into the architecture section, but found that it did not flow right. So this remains as it was in the paper. I added more detail to the section describing the advantages of using perl for doing grid related work. Formatting issues: Added an abstract/summary. Added phone/fax/address contact information. Attached all diagrams in original format as separate files.