Subject: Re: C465: The Software Architecture of a Distributed Problem-SolvingEnvironment Resent-Date: Fri, 21 Jul 2000 11:40:09 -0400 Resent-From: Geoffrey Fox Resent-To: Geoffrey Fox Date: Tue, 11 Jul 2000 05:58:53 -0400 From: "David W. Walker" Organization: Cardiff University To: fox@csit.fsu.edu CC: omer , "li.maozhen" , "m.s.shields" , coral@cs.utk.edu, David.W.Walker@cs.cf.ac.uk Dear Geoffrey, I am pleased to attach the PostScipt (cpande-final.ps) of the revised version of our CP&E paper "The Software Architecture of a Distributed Problem Solving Environment" (C465). I have also attached the LaTex files (cpande-final.tex which inputs paper-final.tex) and the figures. Please let me know if you need the paper in some other format. We are grateful for the comments and suggestions of the reviewers which have helped us improve the paper - please pass on our thanks to them if the opportunity arises. Please find below our response to the issues raised by the reviewers. Best Regards David Reviewer 1 ======== We have fixed the typos. We have included a new section 7.2 which discusses our ideas on PSE security, fault tolerance, QoS, debugging, and profiling, as suggested by the reviewer. Reviewer 2 ======== We have included an additional paragraph following the first paragraph of section 2.2. This new paragraph make reference to other visual programming environments (including AVS) with similar basic functionality to the VCCE that we present, and points out some of the important features that distinguishes our work from these other systems. These difference are further elaborated upon later in section 2.2. Reviewer 3 ======== The reviewer suggests we distinguish more clearly between a PSE for an application and a system for building such a PSE. We believe this distinction has been made in section 4.3 where we point ouit that some parts of the PSE architecture we describe are generic and can be used to build PSEs for a number of application areas, and some parts are specific to a particular application area. A completed PSE will always be application-specific, as stated in the first sentence of the introduction, and will be built out of specific and generic parts. The reviewer further comments on the modest nature of the prototypes we present in section 5. We do now have more advanced PSE-based implementations and we have added references to our work that will appear in journals and at conferences over the next 6 months. We have not inserted these into the paper as they would lengthen it, and it was our intention in this paper to focus on the software architecture rather than on implementation. In response to this reviewer's specific comments: 1. We have added a couple of sentences about how a component is scheduled is there is no prefemance model available for it (property 7, page 3). 2. Property 8. A component's purpose is obviously known to the person who developed it, but it may not be known to a potential end-user who is seeking to determine the best way to solve his/her problem. 3. Page 3, paragraph 1 (after list). We have added a reference to the use of compute servers. 4. Page 5, line 7. We have clarified what is meant by a "physical constraint". 5. Page 12, line 1. We describe some simple implementations if section 5, and give references to our other papers that describe our PSE implementation work in more detail. Section 1 contains a forward reference to section 5. 6. Page 13, line ("our PSE..."). This has been reworded to make it clearer. 7. Page 14, line 16. Figure 5 and associated text explains a simple algorithm for determining the "appropriate granularity" for resolving a component. This explains that if a task graph cannot be further refined then it must be accepted as is. 8. Page 17, lines 17-22. Concerning the issue of scheduling using neural networks: we have added a sentence to stress that this is an area that requires future research. 9. Section 4.1. Electronic notebooks are important for certain application areas and uses, especially collaborative projects. The inclusion of section 4.1 indictates that we view them as important. Since we do not have much research experience in this area we have not made a point of forcefully advocating their use. 10. On CORBA's usefullness. Certainly until a couple of years ago CORBA implementations were often flakey and incomplete. Recently high quality CORBA implementations have become available from companies such as Iona and ObjectSpace and these problems have lessened. We have successfully used CORBA for our implementations as discussed in this and other papers. 11. We have included a reference to the recent book by Houstis et al. --------------------------------------------------------------------- Name: vpce-fig.eps vpce-fig.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: steering.eps steering.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: compmodel.eps compmodel.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: irms-dww.eps irms-dww.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: scheduler.eps scheduler.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: mds-pse.eps mds-pse.eps Type: Postscript Document (application/postscript) Encoding: 7bit Name: cpande-final.ps cpande-final.ps Type: Postscript Document (application/postscript) Encoding: 7bit Name: cpande-final.tex cpande-final.tex Type: TeX Document (application/x-tex) Encoding: 7bit Name: paper-final.tex paper-final.tex Type: TeX Document (application/x-tex) Encoding: 7bit David W. Walker Dr Oak Ridge National Laboratory Computer Science and Mathematics Division David W. Walker Dr Oak Ridge National Laboratory HTML Mail Computer Science and Mathematics Division P. O. Box 2008 Building 6012 Fax: +1 865 574 0680 Oak Ridge Work: +1 865 241-4624 Tennessee 37831-6367 USA Additional Information: Last Name Walker First Name David Version 2.1