sirable, while others might not see that feature as important.

Standards. Another topic of disagreement concerns the importance of standardized specifications. Unfortunately (but not uncommonly in the world of Web services), we are faced with not one but two sets of Web services specifications, similar in purpose and design but different in minor aspects of syntax and semantics. WS-RF has been submitted to, and reviewed and approved by, a standards body ( OASIS); WS-Transfer has been submitted to W3C but is not yet a W3C recommendation. The proposed consolidation of the two, WS-ResourceTransfer, adds to the confusion. Thus, we get a mix of opinions, including the following:

˲ WS-RF, having undergone intensive community review by a standards body, is therefore technically superior and/or morally preferable to the “ proprietary” WS-Transfer.

˲ WS-RF is superior to WS-Transfer for technical reasons—for example, its support for access to subsets of a resource’s state, which can be important if that state is large. (In WS-Transfer, the same effect can be achieved, but only by defining an auxiliary operation that returns an EPR to a desired subset.)

˲ WS-Transfer is superior to WS-RF for technical reasons—for example, its greater simplicity.

˲ As a general principle, we should employ only specifications that are stable, widely accepted, and supported by interoperable tooling. Because neither WS-RF nor WS-Transfer is supported by all major vendors, neither passes this test. This view argues for the no-conventions approach.

˲ HTTP is widely used on the Web and, as a result, it should be preferred over any WS-based solution.

Implementation reuse. Proponents of conventions such as WS-RF and WS-Transfer argue that the adoption of conventional patterns can facilitate code reuse. Every WS-RF or WS-Transfer service performs such things as state access, lifetime management, concurrency control for incoming requests, and state activation/deactiva-tion in the same way. Thus, the code that implements those behaviors can be reused in service implementations.

Proponents of the no-conventions

approach, however, reply that service implementers can also use the same code. In other words, there is certainly value to providing service implementers with standardized implementations of common tasks, but this need not imply that those patterns are exposed outside the service.

Simplicity vs. structure. Proponents of the HTTP/REST approach emphasize that it provides for more concise requests and permits the use of simpler client tooling than approaches based on Web services. Critics point out that the use of HTTP/REST means that users cannot leverage the significant investment in Web services technologies and platforms for message-based interactions.

summary

We have presented four different approaches to modeling state in (Web) service interactions. Each approach defines roughly comparable constructs for referring to, accessing, and managing state components, but differs according to both its precise syntax and the use made (or not) of conventional domain-independent encodings of operations.

Thus, when defining state management operations, the WS-RF and WS-Transfer approaches both use EPRs to refer to state components and to adopt conventions defined in the WS-RF and related specifications and in the WS-Transfer and related specifications, respectively. In contrast, the no-conventions and REST approaches adopt domain-specific encodings of operations, on top of SOAP and HTTP, respectively.

Analysis of the debates that have occurred around these different approaches emphasizes the difficulties inherent in separating technical, political, and stylistic concerns. Some differences of opinion relate to well-defined technical issues and reflect either different philosophies concerning system design or different target applications. Others relate to differing target time scales. For example, no-conventions proponents initially argued against the use of WS-Addressing because of lack of support for that specification in certain tools, while WS-RF and WS-Transfer proponents argued in favor, believing that WS-Addressing would

eventually become universal. Support for WS-Addressing has since become quasi-universal, and now few find its use objectionable.

acknowledgments

We are grateful to Karl Czajkowski, Jim Gray, and Sam Meder for comments on an earlier version of this document. Needless to say, the characterizations of the different arguments are our own. The work of the first author was supported in part by the Mathematical, Information, and Computational Sciences Division subprogram of the Office of Advanced Scientific Computing Research, U.S. Department of Energy, under Contract W-31-109-Eng- 38. The work at Newcastle was supported by the UK eScience Programme, with funding from the EPSRC, DTI and JISC.

 

References

1. Booth, D., Haas, H., McCabe, F., Newcomer, E., Champion, M., Ferris, C. and Orchard, D. Web Services Architecture, W3C, 2003.

2. Chervenak, A., Schopf, J.M., Pearlman, L., Su, M.-H., Bharathi, S., Cinquini, L., D’Arcy, M., Miller, N. and Bernholdt, D. Monitoring the Earth System Grid with MDS4. 2nd IEEE Intl. Conference on e-Science and Grid Computing. Amsterdam, Netherlands, 2006.

3. Fielding, R. Architectural styles and the design of network-based soft ware architectures. Information and Computer Science, University of California Irvine, 2000.

4. Foster, I., Czajkowski, K., Ferguson, D., Frey, J., Graham, S., Maguire, T., Snelling, D. and Tuecke, S. Modeling and managing state in distributed systems: The role of OGSI and WSRF. Proceedings of the IEEE 93, 3. 604–612.

5. Helland, P. Data on the Inside vs. Data on the Outside. Microsoft Corporation, 2004.

6. Parastatidis, S., Webber, J., Watson, P. and Rischbeck, T. WS-GAF: A framework for building grid applications using Web Services. Concurrency and Computation: Practice and Experience 17, 2–4, 391–417.

 

Ian Foster ( [e-mail]) is the director of the Computation Institute at Argonne National Laboratory, where he is also an Argonne Distinguished Fellow, and the University of Chicago, where he is also the Arthur Holly Compton Distinguished Service Professor of Computer Science.

Savas Parastatidis ( [e-mail]) is an architect in Microsoft Research. He investigates the use of technology in eResearch and is particularly interested in Cloud Computing, knowledge representation and management, and social networking.

Paul Watson (Paul. [e-mail]) is a professor of Computer Science at Newcastle University, and Director of the North East Regional e-Science Centre in the U. K.

Mark McKeown was a grid architect at the University of Manchester, U.K., at the time of this work.