Technical Program

This year's Technical Papers track comprises 26 presentations of the latest, top-rate academic and industrial research results. (Please note that, in an unusual move, we have arranged for one of the Technical Papers to be presented as part of the Onward! track because of a natural affinity with one of the Onward! presentations.)

Tuesday, 28 October


Onward! Free Culture: The limited but essential role of property in building an environment for creativity


Technical Paper: Refactoring and Reflection
Panel: Meeting the Challenge of Software Engineering Education for Working Professionals in the 21st Century
Onward!: Naturalistic Programming and Run-Time Code Generation


Keynotes & Invited Speakers: Grid-supported Communications Applications


Panel: Xtreme Programming and Agile Coaching
Practitioners Report: Tools


Technical Paper: Smalltalkiana

Wednesday, 29 October


Keynotes & Invited Speakers: Seven Paradoxes of Object-Oriented Programming Languages


Technical Paper Technical Papers and Onward!: Error Repair
Technical Paper: Generics
Panel: Discipline and Practices of TDD (Test Driven Development)
Onward!: Acceptability-Oriented Computing
Domain-Driven Development: Vision


Technical Paper: Java Performance
Panel: Innovate!
Practitioners Report: Old Code
Domain-Driven Development: Technology


Technical Paper: Language Design
Panel Model Driven Architecture: How far have we come, how far can we go?
Practitioners Report: Real-World Objects
Onward!: The Biology of Information

Thursday, 30 October


Keynotes & Invited Speakers: OOPSLA Keynote - The Internet Paradigm Shift
Technical Paper: Garbage Collection 1
Panel: Agile Management—An Oxymoron?


Technical Paper: Garbage Collection 2
Panel Object-Oriented Success Stories: "Learning from our Failures"
Onward! Panel: Reuse Repositories and Reuse—The Realities


Technical Paper: Analysis


Keynotes & Invited Speakers: Eclipse and The Dark Side of the Moon
Technical Paper: Transactions and Persistence
Panel: What's so eXtreme About Doing Things Right?
Panel Application Servers: One Size Fits All ... Not?

Free Culture: The limited but essential role of property in building an environment for creativity

Tuesday, 28 October – 8:30-10:00

Larry Lessig, Stanford University,

This talk will sketch the boundaries of protection that intellectual property law should set, but argue that extremism has now defeated these limits. The consequence is an environment within which modular creativity is increasingly constrained.


Lawrence Lessig is a Professor of Law at Stanford Law School and founder of the school's Center for Internet and Society. He is one of the country's leading commentators on legal aspects of new communications technologies and cyberspace. He is Professor of Law and founder and executive director of the Center for Internet and Society at Stanford Law School. Professor Lessig teaches and writes in the areas of constitutional law, law and high technology, Internet regulation, comparative constitutional law, and the law of cyberspace.

He is the author of many influential publications about cyberlaw and cyberspace, including two books: The Future of Ideas: The Fate of the Commons in a Connected World (2001) and Code and Other Laws of Cyberspace (1999).

Refactoring and Reflection

Tuesday, 28 October – 10:30-12:00

10:30 - 11:00
Language-Independent Aspect-Oriented Programming

Donal Lafferty, Trinity College Dublin,
Vinny Cahill, Trinity College Dublin,

The term aspect-oriented programming (AOP) has come to describe the set of programming mechanisms developed specifically to express crosscutting concerns. Since crosscutting concerns cannot be properly modularized within object-oriented programming (OOP), they are expressed as aspects and are composed, or woven, with traditionally encapsulated functionality referred to as components.

Many AOP models exist, but their implementations are typically coupled with a single language. To allow weaving of existing components with aspects written in the language of choice, AOP requires a language-independent tool.

This paper presents Weave.NET, a load-time weaver that allows aspects and components to be written in a variety of languages and freely intermixed. Weave.NET relies on XML to specify aspect bindings and ECMA Common Language Infrastructure to avoid coupling aspects or components with a particular language.

By demonstrating language-independence, Weave.NET provides a migration path to the AOP paradigm by preserving existing developer knowledge, tools and software components. The tool's capabilities are demonstrated with logging aspects written with and applied to Visual Basic and C# components.

11:00 - 11:30
Refactoring for Generalization Using Type Constraints

Frank Tip, IBM T.J. Watson Research Center,
Adam Kiezun, IBM Research OTI Labs,
Dirk Baeumer, IBM Research OTI Labs,

Refactoring is the process of applying behavior-preserving transformations (called "refactorings") in order to improve a program's design. Associated with a refactoring is a set of preconditions that must be satisfied to guarantee that program behavior is preserved, and a set of source code modifications. An important category of refactorings is concerned with generalization (e.g., "Extract Interface" for re-routing the access to a class via a newly created interface, and "Pull Up Members" for moving members into a superclass). For these refactorings, both the preconditions and the set of allowable source code modifications depend on interprocedural relationships between types of variables. We present an approach in which type constraints are used to verify the preconditions and to determine the allowable source code modifications for a number of generalization-related refactorings. This work is implemented in the standard distribution of Eclipse (see

11:30 - 12:00
Partial Behavioral Reflection: Spatial and Temporal Selection of Reification

Eric Tanter, University of Chile, École des Mines de Nantes/INRIA,
Jacques Noyé, École des Mines de Nantes/INRIA,
Denis Caromel, Université de Nice,
Pierre Cointe, École des Mines de Nantes/INRIA,

Behavioral reflection is a powerful approach for adapting the behavior of running applications. In this paper we present and motivate partial behavioral reflection, an approach to more efficient and flexible behavioral reflection. We expose the spatial and temporal dimensions of such reflection. In the context of Java, we present a reflective architecture offering appropriate interfaces for static and dynamic configuration of partial behavioral reflection at various levels, as well as an open reflective extension for Java implementing this architecture. Reflex is the first extension that fully supports partial behavioral reflection in a portable manner, and that seamlessly integrates load-time and runtime behavioral reflection, along with static optimizations. The paper shows preliminary benchmarks and examples supporting the approach. The examples, dealing with the observer pattern and asynchronous communication via transparent futures, also show the interest of partial behavioral reflection as a tool for open dynamic aspect-oriented programming.

Meeting the Challenge of Software Engineering Education for Working Professionals in the 21st Century

Tuesday, 28 October – 10:30-12:00

Steven Fraser (Chair), Independent Consultant,
Ray Bareiss, CMU (West),
Barry Boehm, USC,
Mark Hayes, Microsoft,
Laura Hill, Sun Microsystems,
Gabby Silberman, IBM,
Dave Thomas, Bedarra Research Labs,

Software engineering education for working professionals remains a challenge from the perspective of determining relevant content; identifying effective methods for delivery; and maintaining the focus and motivation of students. This panel brings together academic and industry professionals to share their perspectives and experiences. Anticipated points for discussion include: education/training delivery strategies, curriculum definition, marketing issues, collaboration strategies to engage industry sponsorship, value assessments for students and sponsoring organizations, and program success stories. This will be a highly interactive panel and the audience should come prepared to both ask and answer questions.

Naturalistic Programming and Run-Time Code Generation

Tuesday, 28 October – 10:30-12:00

These papers explore simplifying programming by examining some of our underlying assumptions about references and constructing programs.

Beyond AOP: Toward Naturalistic Programming

Cristina Lopes, University of California, Irvine,
Paul Dourish, University of California, Irvine,
David Lorenz, Northeastern University,
Karl Lieberherr, Northeastern University,

Software understanding (for documentation, maintenance or evolution) is one of the longest-standing problems in Computer Science. The use of bct-oriented languages helps, but fundamentally remains far from solving the problem. Most programming languages and systems have fallen prey to the assumption that they are supposed to capture idealized models of computation inspired by deceptively simple metaphors such as objects and mathematical functions. Aspect-oriented programming languages have made a significant breakthrough by noticing that, in many situations, humans think and describe in crosscutting terms. In this paper we suggest that the next breakthrough would require looking even closer to the way humans have been thinking and describing complex systems for thousand of years using natural languages. While natural languages themselves are not appropriate for programming, they contain a number of elements that make descriptions concise, effective and understandable. In particular, natural languages referentiality is a key factor in supporting powerful program organizations that can be more easily understood by humans.

Routine Run-time Code Generation

Sam Kamin, Univ. of Illinois at Urbana-Champaign,

The decades-old paradigm of "object code = executable code: is outdated. The paradigm we propose in this paper is "object code = executable program generator.: Similarly, "software component" should be "program that, when appropriately invoked by a client, generates machine language useful to the client." Traditional object files amount to constant functions in the domain of code-producing functions; in the future, we will routinely employ the full domain of higher-order code values. In this paper, we argue that such a change in viewpoint can lead to much wider use of run-time code generation, by overcoming the bureaucratic barriers it usually encounters. Run-time code generation has numerous applications; when it comes to be used routinely, many more applications are likely to be discovered.

Grid-supported Communications Applications

Tuesday, 28 October – 13:30-15:00

Gerald Labedz, Motorola Labs,

This invited talk will describe how emerging distributed computing systems could mutate into the communications systems of the future. Armed with high speed interconnect, and software elements that assign, track and charge for geographically distributed computing resources, these systems could just as easily compute and deliver an image to a remote location (for human communications) as perform an internal streaming file transfer. With this kind of capability not tied to the user's premises, customer equipment could be very asset light and still deliver very high end services like realtime processed video. Using as an example MIRAGE II, a data mining system with a geographically distributed, remotely computed Augmented Reality interface, a prototype built by Motorola in association with NCSA, Labedz will discuss the main hurdles along the way to communications systems dependent on real-time remote computing and high speed networks.


Gerry Labedz has worked at Motorola for 28 years in the areas of mobile and cellular radio hardware and software. He worked on the first microprocessor controlled radio in the '70s, the first digital cellular system in the '80s, and the world's most accurate parallel-computed cellular system simulation tool in the '90s. He holds 23 patents in the areas of radio signaling, system design and visualization, and large-scale real-time simulation for cellular system control. During the 1990s, he and his group were the largest industrial users of the supercomputer facilities at the National Center for Supercomputing Applications (NCSA). He holds the title of Dan Noble Fellow, Motorola's highest technical honor. Most recently he has led a project centering on Hyperfast Computational Communications, or HyperComputiCations

Xtreme Programming and Agile Coaching

Tuesday, 28 October – 15:30-17:00

Steven Fraser (Impresario), Independent Consultant,
Rachel Reinitz (Chair), IBM,
Jutta Eckstein, Independent Consultant,
Joshua Kerievsky, Industrial Logic,
Rob Mee, Pivotal Computer Systems,
Mary Poppendieck, Agile Alliance,

This panel brings together coaches to discuss all aspects of the practice: how to become a coach, choosing a coach, and describing what is to be an (in) effective coach. A coach watches, provides feedback, and suggests subtle direction. The coach may be more—for example—an architect or team lead—but that is a matter for debate. This session will be run as a panel with two open "fish bowl" seats, only one of which may be occupied by audience members at any one time. The panelists will defend their positions and offer feedback. Panelists were asked to offer their observations on three questions: How did YOU become a coach? What's the toughest thing you've had to do as a coach? What's your advice for teams looking for a coach?


Tuesday, 28 October – 15:30-17:00

15:30 - 16:00
Experience in Developing the UrbanSim System: Tools for Developing Research Software for a Politically Charged Domain

Bjorn Freeman-Benson, Center for Urban Simulation and Policy Analysis, University of Washington,
Alan Borning, Department of Computer Science and Engineering, University of Washington,

UrbanSim is a sophisticated simulation system for modeling urban land use, transportation, and environmental impacts, over a period of 20 or more years, under alternate possible scenarios. Our purpose is to support democratic decision-making on such issues as building new transit systems or freeways, or changing zoning or economic incentives, as well as on broader issues such as sustainable, livable cities, economic vitality, social equity, and environmental preservation. The domain is politically charged, with different stakeholders bringing strongly held values to the table. Our goal is not to favor any particular stakeholder values, but rather to let the stakeholders evaluate the results themselves. This need for neutrality has several implications for the development process, among other things resulting in our choice of an open source/open process model. UrbanSim is also an NSF-funded infrastructure project with a mix of research and development goals. Unlike most academic projects, this one has real customers (municipal planning organizations) with real needs.

Thus the UrbanSim project has a number of simultaneous goals, which are in some tension: producing credible, reliable code; utilizing a flexible process that adapts to new research discoveries; maintaining coherent architecture and continuous development progress; and using part-time and inexperienced student staff in addition to experienced full-time staff. (This is in contrast to the usual model in XP and other agile development processes, in which all developers are considered equally experienced and skilled.)

In this report, we discuss the tools we have chosen, modified, or developed in-house; and the reasons why these tools are good choices for this mixed research/product development environment. Our tools include: Eclipse, MySQL, CVS, Ant, Fireman (a Tinderbox-like automatic build and test system), JUnit, FIT, Wiki, Bugzilla, a physical traffic light, and a daily project dashboard. The tools are tightly integrated and collectively support our agile process.

16:00 - 16:30
Experiences using an ODBMS for a high-volume internet banking system

Vincent Coetzee,,
Robert Walker, GemStone Systems,

Few large corporate organizations make the decision to use an Object Database Management System (ODBMS) when developing high volume transactional eCommerce web sites. This report examines the architecture used to power a website that encompasses banking, online shopping, and the management of a Customer Loyalty Currency called eBucks. This system proves that an ODBMS can be used in a high volume transactional system with great success. While the choice of this technology has many merits, there are drawbacks. These drawbacks are examined along with the solutions that have been used at eBucks to either solve or ameliorate them.


Tuesday, 28 October – 15:30-16:30

15:30 - 16:00
Applying Traits to the Smalltalk Collection Classes

Andrew P. Black, OGI School of Science & Engineering, Oregon Health & Science University,
Nathanael Schärli, University of Bern,
Stéphane Ducasse, University of Bern,

Traits are a programming language technology modeled after mixins but avoiding their problems. This paper reports on a refactoring of the Smalltalk collections classes using traits. We observed that the original collection classes contained much duplication of code; traits let us remove all of it. We also found places where the protocols of the collections lacked uniformity; traits allow us to correct these non-uniformities without code duplication. In addition, traits make possible more general reuse of collection code outside of the existing hierarchy; for example, they make it easy to convert other collection-like things into true collections. Our refactoring reduced the number of methods in the collection classes by approximately 10 per cent. More importantly, understandability and reusability of the code was significantly improved.

16:00 - 16:30
OOPAL: Integrating Array Programming in Object-Oriented Programming

Philippe Mougin,
Stéphane Ducasse, University Of Bern,

Array programming shines in its ability to express computations at a high-level of abstraction, allowing one to manipulate and query whole sets of data at once. This paper presents the OOPAL model that enhances object-oriented programming with array programming features. The goal of OOPAL is to determine a minimum set of modifications that must be made to the traditional object model in order to take advantage of the possibilities of array programming. It is based on a minimal extension of method invocation and the definition of a kernel of methods implementing the fundamental array programming operations. The model is validated in F-Script, a new scripting language.

Seven Paradoxes of Object-Oriented Programming Languages

Wednesday, 29 October – 8:30-10:00

David Ungar, Sun Microsystems

Although many of us have worked to create good object-oriented programming languages, it would be hard to say (with a straight face) that any of our creations have totally succeeded. Why not? I believe that this endeavor is essentially paradoxical. Thus, whenever a language designer pursues a particular goal and loses sight of the lurking paradox, the outcome is an all too often fatally flawed result. One way to think about this is to explore the following seven paradoxes:

    1. Because programming languages, development environments, and execution engines are intended for both people and computers, they must both humanize and dehumanize us.
    2. Adding a richer set of concepts to a programming language impoverishes its universe of discourse.
    3. Putting a language's cognitive center in a more dynamic place reduces the verbiage needed to accomplish a task, even though less information can be mechanically deduced about the program.
    4. The most concrete notions are the most abstract, and pursuing comfort or correctness with precision leads to fuzziness.
    5. Although a language, environment, and execution engine are designed for the users' minds, the experience of use will alter the users' minds.
    6. Object-oriented programming has its roots in modeling and reuse, yet these notions do not coincide and even conflict with each other.
    7. A language designed to give programmers what they want may initially succeed but create pernicious problems as it catches on. However, a language designed to give programmers what they really need may never catch fire at all.

Many of these assertions seem nonsensical, misguided, or just plain wrong. Yet, a deeper understanding of these paradoxes can point the way to better designs for object-oriented programming languages.

Technical Papers and Onward!: Error Repair

Wednesday, 29 October – 10:30-12:00

11:30 - 12:00
Automatic Detection and Repair of Errors in Data Structures

Brian Demsky, MIT Laboratory for Computer Science,
Martin Rinard, MIT Laboratory for Computer Science,

We present a system that accepts a specification of key data structure constraints, then dynamically detects and repairs violations of these constraints, enabling the program to continue to execute productively even in the face of otherwise crippling errors. Our experience using our system indicates that the specifications are relatively easy to develop once one understands the data structures. Furthermore, for our set of benchmark applications, our system can effectively repair errors to deliver consistent data structures that allow the program to continue to operate successfully within its designed operating envelope.


Wednesday, 29 October – 10:30-12:00

10:30 - 11:00
A First-Class Approach to Genericity

Eric Allen, Rice University,
Jonathan Bannet, Rice University,
Robert Cartwright, Rice University,

This paper describes how to add first class generic types—including mixins—to strongly-typed, object-oriented languages with nominal (name-based) subtyping such as Java and C#. A generic type system is "first-class" if generic types can appear in any context where conventional types can appear. In this context, a mixin is simply a generic class that extends one its type parameters, e.g., a class C<T> that extends T. Although mixins of this form are widely used in C++ (via templates), they are clumsy and error-prone because C++ treats mixins as macros, forcing each mixin instantiation to be separately compiled and type-checked. The abstraction embodied in a mixin is never separately analyzed.

Our formulation of mixins using first class genericity accommodates sound local (class-by-class) type checking, in the same sense that Java supports local (class-by-class) compilation. A mixin can be fully type-checked given symbol tables for each of the classes that it directly references—the same context in which Java performs incremental class compilation. To our knowledge no previous formal analysis of first-class genericity in languages with nominal type systems has been conducted, which is surprising because nominal type systems have become predominant in mainstream object-oriented programming languages.

What makes our treatment of first class genericity particularly interesting and important is the fact that it can be added to the existing Java language without any change to the underlying Java Virtual Machine. Moreover, the extension is backward compatible with legacy Java source and class files. Although our discussion of a practical implementation strategy focuses on Java, the same scheme could be applied to other object-oriented languages such as C# or Eiffel that support incremental compilation, dynamic class loading, and a static type system with nominal subtyping.

11:00 - 11:30
A Comparative Study of Language Support for Generic Programming

Ronald Garcia, Indiana University,
Jaakko Jarvi, Indiana University,
Andrew Lumsdaine, Indiana University,
Jeremy Siek, Indiana University,
Jeremiah Willcock, Indiana University,

Many modern programming languages support basic generic programming, sufficient to implement type-safe polymorphic containers. Some languages have moved beyond this basic support to a broader, more powerful interpretation of generic programming, and their extensions have proven valuable in practice. This paper reports on a comprehensive comparison of generics in six programming languages: C++, Standard ML, Haskell, Eiffel, Java (with its proposed generics extension), and Generic C#. By implementing a substantial example in each of these languages, we identify eight language features that support this broader view of generic programming. We find these features are necessary to avoid awkward designs, poor maintainability, unnecessary run-time checks, and painfully verbose code. As languages increasingly support generics, it is important that language designers understand the features necessary to provide powerful generics and that their absence causes serious difficulties for programmers.

11:30 - 12:00
Lightweight Confinement for Featherweight Java

Tian Zhao, University of Wisconsin, Milwaukee,
Jens Palsberg, Purdue University,
Jan Vitek, Purdue University,

Confinement properties impose a structure on object graphs which can be used to enforce encapsulation— which is essential to certain program optimizations, to modular reasoning, and in many cases to software assurance. This paper formalizes the notion of confined type in the context of Featherweight Java. A static type system that mirrors the informal rules of Vitek and Bokopwski is proposed and proven sound. The definition of confined types is extended to confined instantiation of generic classes. Thus allowing for confined collection types in Java and for classes that can be confined post hoc. Confinement types rules are given for Generic Featherweight Java, and proven sound.

Discipline and Practices of TDD (Test Driven Development)

Wednesday, 29 October – 10:30-12:00

Steven Fraser (Chair), Independent Consultant,
Dave Astels, Adaption Software,
Kent Beck, Three Rivers Institute,
Barry Boehm, USC,
John McGregor, Clemson University,
James Newkirk, Microsoft,
Charlie Poole, Poole Consulting,

This panel brings together practitioners with experience in Agile and XP methodologies to discuss the approaches and benefits of applying Test Driven Development (TDD). The goal of TDD is clean code that works. The mantra of TDD is: write a test; make it run; and make it right. Open questions to be addressed by the panel include:

  • How are TDD approaches to be applied to databases, GUIs, and distributed systems?
  • What are the quantitative benchmarks that can demonstrate the value of TDD, and
  • What are the best approaches to solve the ubiquitous issue of scalability?

Acceptability-Oriented Computing

Wednesday, 29 October – 10:30-12:00

These papers talk about a new approach to constructing software systems in which the designer identifies key properties that the execution must satisfy to be acceptable to its users. The resulting system consists of layers of components enforcing the acceptability properties.

Acceptability-Oriented Computing

Martin Rinard, MIT,

We propose a new approach to the construction of software systems. Instead of attempting to build a system that is as free of errors as possible, we instead propose that the designer identify key properties that the execution must satisfy to be acceptable to its users. The developer then augments the system with a layered set of components, each of which enforces one of the acceptability properties. The potential advantages of this approach include more flexible, resilient systems that recover from errors and behave acceptably across a wide range of operating environments, an appropriately prioritized investment of engineering resources, and reduced development costs because of the ability to productively incorporate unreliable components into the final software system and to use less skilled implementors in the development process.

Automatic Detection and Repair of Errors in Data Structures

Brian Demsky, MIT,
Martin Rinard, MIT,

We present a system that accepts a specification of key data structure constraints, then dynamically detects and repairs violations of these constraints, enabling the program to continue to execute productively even in the face of otherwise crippling errors. Our experience using our system indicates that the specifications are relatively easy to develop once one understands the data structures. Furthermore, for our set of benchmark applications, our system can effectively repair errors to deliver consistent data structures that allow the program to continue to operate successfully within its designed operating envelope.


Wednesday, 29 October – 10:30-12:00

Chair: Krzysztof Czarnecki, University of Waterloo,

10:30 - 11:00
Model Driven Development - The Case for Domain Oriented Programming

Dave Thomas, Bedarra Research Labs,
Brian Barry, Bedarra Research Labs,

In this paper, we offer an alternative vision for domain driven development (3D). Our approach is model driven and emphasizes the use of generic and specific domain oriented programming (DOP) languages. DOP uses strong specific languages, which directly incorporate domain abstractions, to allow knowledgeable end users to succinctly express their needs in the form of an application computation. Most domain driven development (3D) approaches and techniques are targeted at professional software engineers and computer scientists. We argue that DOP offers a promising alternative. Specifically we are focused on empowering application developers who have extensive domain knowledge as well as sound foundations in their professions, but may not be formally trained in computer science. We provide a brief survey of DOP experiences, which show that many of the best practices such as patterns, refactoring, and pair programming are naturally and ideally practiced in a MDD setting. We compare and contrast DOP with other popular approaches, most of which are deeply rooted in the OO community. Finally we highlight challenges and opportunities in the design and implementation of such languages.

11:00 - 11:30
An End to End Domain Driven Developement Framework

Aditya Agrawal, ISIS, Vanderbilt University,
Gabor Karsai, ISIS, Vanderbilt University,
Akos Ledeczi, ISIS, Vanderbilt University,

This paper presents a comprehensive, domain-driven framework for development. It consists of a meta-programmable domain-specific modeling environment, and a model transformation and generation toolset based on graph transformations. The framework allows the creation of custom, domain-oriented programming environments that support end-user programmability. In addition, the framework could be considered an early, end-to-end implementation of the concepts advocated by the Model-Driven Architecture of OMG.

11:30 - 12:00
Software Factories: Assembling Applications With Patterns, Models, Frameworks and Tools

Jack J. Greenfield, Microsoft, Visual Studio, Enterprise Frameworks And Tools,
Keith W. Short, Microsoft, Visual Studio, Enterprise Frameworks And Tools,

The confluence of component based development, model driven development and software product lines forms an approach to application development based on the concept of software factories. This approach promises greater gains in productivity and predictability than those produced by the incremental advances of recent memory, which have not kept pace with rapid innovation in platform technology. Focusing on development by assembly using domain specific languages, patterns, models, frameworks and tools, software factories make systematic reuse cost effective for many applications, enabling the formation of supply chains and opening the door to mass customization.

Java Performance

Wednesday, 29 October – 13:30-15:00

13:30 - 14:00
Dynamic Metrics for Java

Bruno Dufour, McGill University,
Karel Driesen, McGill University,
Laurie Hendren, McGill University,
Clark Verbrugge, McGill University,

In order to perform meaningful experiments in optimizing compilation and run-time system design, researchers usually rely on a suite of benchmark programs of interest to the optimization technique under consideration. Programs are described as numeric, memory-intensive, concurrent, or object-oriented, based on a qualitative appraisal, in some cases with little justification. We believe it is beneficial to quantify the behavior of programs with a concise and precisely defined set of metrics, in order to make these intuitive notions of program behavior more concrete and subject to experimental validation. We therefore define a set of unambiguous, dynamic, robust and architecture-independent metrics that can be used to categorize programs according to their dynamic behavior in five areas: size, data structure, memory use, concurrency, and polymorphism. A framework computing some of these metrics for Java programs is presented along with specific results.

14:00 - 14:30
How Java Programs Interact with Virtual Machines at the Microarchitectural Level

Lieven Eeckhout, Ghent University,
Andy Georges, Ghent University,
Koen De Bosschere, Ghent University,

Java workloads are becoming increasingly prominent on various platforms ranging from embedded systems, over general-purpose computers to high-end servers. Understanding the implications of all the aspects involved when running Java workloads, is thus extremely important during the design of a system that will run such workloads, to meet its design goals. In other words, understanding the interaction between the Java application, its input and the virtual machine it runs on, is key to a successful design. The goal of this paper is to study this complex interaction at the microarchitectural level, e.g., by analyzing the branch behavior, the cache behavior, etc. This is done by measuring a large number of performance characteristics using performance counters on an AMD K7 Duron microprocessor. These performance characteristics are measured for seven virtual machine configurations, and a collection of Java benchmarks with corresponding inputs coming from the SPECjvm98 benchmark suite, the SPECjbb2000 benchmark suite, the Java Grande Forum benchmark suite and an open-source raytracer, called Raja with 19 scene descriptions. This large amount of data is further analyzed using statistical data analysis techniques, namely principal components analysis and cluster analysis. These techniques provide useful insights in an understandable way.

From our experiments, we conclude that (i) the behavior observed at the microarchitectural level is primarily determined by the virtual machine for small input sets, e.g., the SPECjvm98 s1 input set; (ii) the behavior can be quite different for various input sets, e.g., short-running versus long-running benchmarks; (iii) for long-running benchmarks with few hot spots, the behavior can be primarily determined by the Java program and not the virtual machine, i.e., all the virtual machines optimize the hot spots to similarly behaving native code; (iv) in general, the behavior of a Java application running on one virtual machine can be significantly different from running on another virtual machine. These conclusions warn researchers working on Java workloads to be careful when using a limited number of Java benchmarks or virtual machines since this might lead to biased conclusions.

14:30 - 15:00
Effectiveness of Cross-Platform Optimizations for a Java Just-In-Time Compiler

Kazuaki Ishizaki, IBM Research, Tokyo Research Laboratory,
Mikio Takeuchi, IBM Research, Tokyo Research Laboratory,
Kiyokuni Kawachiya, IBM Research, Tokyo Research Laboratory,
Toshio Suganuma, IBM Research, Tokyo Research Laboratory,
Osamu Gohda, IBM Research, Tokyo Research Laboratory,
Tatsushi Inagaki, IBM Research, Tokyo Research Laboratory,
Akira Koseki, IBM Research, Tokyo Research Laboratory,
Kazunori Ogata, IBM Research, Tokyo Research Laboratory,
Motohiro Kawahito, IBM Research, Tokyo Research Laboratory,
Toshiaki Yasue, IBM Research, Tokyo Research Laboratory,
Takeshi Ogasawara, IBM Research, Tokyo Research Laboratory,
Tamiya Onodera, IBM Research, Tokyo Research Laboratory,
Hideaki Komatsu, IBM Research, Tokyo Research Laboratory,
Toshio Nakatani, IBM Research, Tokyo Research Laboratory,

We describe the system overview of our Java JIT compiler, which has been the basis for the latest production version of IBM Java virtual machine that supports a diversity of processor architectures including both 32-bit and 64-bit modes, CISC, RISC, and VLIW architectures. In particular, we focus on the design and evaluation of the cross-platform optimizations that are common across different architectures. We study the effectiveness of each optimization by selectively disabling it in our JIT compiler on three different platforms: IA32, IA64, and PowerPC. Based on the detailed statistics, we classify our optimizations and identify a small set of the most cost-effective ones in terms of the performance improvement as the benefit and the compilation time as the cost. In summary, we demonstrate that, with a selected set of optimizations, we can achieve 90% of the peak performance for SPECjvm98 at the expense of only 33% of the total compilation time in comparison to the case in which all the optimizations are enabled.


Wednesday, 29 October – 13:30-15:00

Laura Hill (Chair), Sun Microsystems, Inc.,
Rachel Davies, Amarinda,
Dick Gabriel, Sun Microsystems, Inc.,
Harlan Sexton, Oracle Corp,
Kevin Tyson, Independent Consultant,
David West, New Mexico Highlands University and University of New Mexico,

Freedom to innovate is one of the key motivators for many technical workers. Unfortunately, although innovation is often trumpeted as a key company attribute, it seems that many organizations struggle to provide the necessary environment—even those organizations whose original claim to fame lay in their ability to innovate. This panel will look at the barriers to innovation that occur in a variety of environments: large, well-established organizations, start-ups, academia, standards bodies and the open source community. Panelists will propose a set of technical and non-technical techniques that can be used to foster innovation in even the most lethargic or hostile environment.

Old Code

Wednesday, 29 October – 13:30-15:00

13:30 - 14:00
Using AspectJ for Component Integration in Middleware

Adrian Colyer, IBM UK,
Ron Bodkin, New Aspects of Security,
Jim Hugunin, PARC,
Andrew Clement, IBM UK,

This report discusses experiences applying and enhancing AspectJ for a middleware product line at IBM. The purpose of this effort was to apply Aspect Oriented Programming to cleanly separate components from their execution platforms, while still allowing them to take advantage of platform-specific facilities for aspects such as error handling, performance monitoring, and logging. It presents the evolution of the design, implementation, tools support, and approaches used to achieve organizational buy in.
Keywords: AspectJ, aspect oriented programming, middleware

14:00 - 14:30
Five years of framework building: lessons learned

Kurt Madsen, MetaTech, Inc.,

When developing large software systems, it is often difficult to foresee exactly which trade-offs are important, and which quality parameters will be of importance down the road. This paper reports experiences from a project in which a large application framework for B2B integration has been continuously developed and used over a five year period. The framework has been the foundation for a variety of different concrete applications; here we will report on our experiences from this endeavor.

14:30 - 15:00
Agile Regression Testing Using Record & Playback

Gerard Meszaros, ClearStream Consulting,
Ralph Bohnet, ClearStream Consulting,
Jennitta Andrea, ClearStream Consulting,

There are times when it is not practical to hand-script automated tests for an existing system before one starts to modify it (whether to refactor it to permit automated testing or to add new functionality). In these circumstances, the use of “record & playback” testing may be a viable alternative to hand-writing all the tests. This paper describes experiences using this approach and summarizes key learnings applicable to other projects.


Wednesday, 29 October – 13:30-15:00

Chair: John Vlissides, IBM T. J. Watson Research Center,

13:30 - 14:00
XAspects: An Extensible System for Domain-Specific Aspect Languages

Macneil Shonle, Northeastern University,
Karl Lieberherr, Northeastern University,
Ankit Shah, Northeastern University,

Current general aspect-oriented programming solutions fall short of helping the problem of separation of concerns for several concern domains. Because of this limitation good solutions for these concern domains do not get used and the opportunity to benefit from separation of these concerns is missed. By using XAspects, a plug-in mechanism for domain-specific aspect languages, separation of concerns can be achieved at a level beyond what is possible for object-oriented programming languages. As a result, XAspects allows for certain domain-specific solutions to be used as easily as a new language feature.
Keywords: Aspect-oriented programming, — programming, language extensions, domain-specific languages.

14:00 - 14:30
The Power of Symmetry: Unifying Inheritance and Generative Programming

DeLesley Hutchins, MZA Associates Corporation,

I present the Ohmu language, a unified object model which allows a number of "advanced" techniques such as aspects, mixin layers, parametric polymorphism, and generative components to be implemented cleanly using two basic concepts: block structure and inheritance. I argue that conventional ways of defining classes and objects have created artificial distinctions which limit their expressiveness. The Ohmu model unifies functions, classes, instances, templates, and even aspects into a single construct – the structure. Function calls, instantiation, aspect-weaving, and inheritance are likewise unified into a single operation – the structure transformation. This simplification eliminates the distinction between classes and instances, and between compile-time and run-time code. Instead of being compiled, programs are reduced using partial evaluation, in which the interpreter is invoked at compile-time. Within this architecture, standard OO inheritance becomes a natural vehicle for creating meta-programs and automatic code generators— the key to a number of recent domain-driven programming methodologies.

14:30 - 15:00
Domain Driven Web Development With WebJinn

Sergei Kojarski, Northeastern University,
David Lorenz, Northeastern University,

Web application development cuts across the HTTP protocol, the client-side presentation language (HTML, XML), the server-side technology (Servlets, JSP, ASP, PHP), and the underlying resource (files, database, information system). Consequently, web development concerns including functionality, presentation, control, and structure cross-cut, leading to tangled and scattered code that is hard to develop, maintain, and reuse. In this paper we analyze the cause, consequence, and remedy for this crosscutting. We distinguish between intra-crosscutting that results in code tangling and inter-crosscutting that results in code scattering. To resolve inter-crosscutting, we present a new web application development model named XP that introduces extension points as place-holders for structure-dependent code. We present another model named DDD that incorporates XP into the Model-View-Controller (MVC) model to resolve both intra- and inter-crosscutting. WebJinn is a novel domain-driven web development framework that implements the DDD model. WebJinn has been used to develop web applications at several web sites. Domain driven web development with WebJinn benefits from a significant improvement in code reuse, adaptability, and maintainability.

Language Design

Wednesday, 29 October – 15:30-17:00

15:30 - 16:00
HydroJ: Object-Oriented Pattern Matching for Evolvable Distributed Systems

Keunwoo Lee, University of Washington,
Anthony LaMarca, Intel Research Seattle,
Craig Chambers, University of Washington,

In an evolving software system, components must be able to change independently while remaining compatible with their peers. One obstacle to independent evolution is the brittle parameter problem}: the ability of two components to communicate can depend on a number of inessential details of the types, structure, and/or contents of the values communicated. If these details change, then the components can no longer communicate, even if the essential parts of the message remain unaffected.

We present HydroJ, an extension of Java that addresses this problem. In HydroJ, components communicate using self-describing, semi-structured messages, and programmers use pattern matching to define the handling of messages. This design stems from two central ideas: first, that self-describing messages reduce dependence on inessential message format details; and second, that object-oriented pattern matching naturally focuses on the essential information in a message and is insensitive to inessential information.

We have developed these ideas in the context of Rain, a distributed, heterogeneous messaging system for ubiquitous computing. To evaluate the design, we have constructed a prototype HydroJ compiler, implemented some Rain services in HydroJ, and formalized HydroJ's key features in a core language.

16:00 - 16:30
Relaxed MultiJava: Balancing Extensibility and Modular Typechecking

Todd Millstein, University of Washington,
Mark Reay, University of Washington,
Craig Chambers, University of Washington,

We present the rationale, design, and implementation of Relaxed MultiJava (RMJ), a backward-compatible extension of Java that allows programmers to add new methods to existing classes and to write multimethods. Previous languages supporting these forms of extensibility either restrict their usage to a limited set of programming idioms that can be modularly typechecked (and modularly compiled) or simply forego modular typechecking altogether. In contrast, RMJ supports the new language features in a virtually unrestricted form while still providing modular static typechecking and compilation. In some cases, the RMJ compiler will warn that the potential for a type error exists, but it will still complete compilation. In that case, a custom class loader transparently performs load-time checking to verify that the potential error is never realized. RMJ's compiler and custom loader cooperate to keep load-time checking costs low. We report on our qualitative and quantitative experience with our implementation of RMJ.

16:30 - 17:00
ModJava: A Rational Module System for Java and Its Applications

John Corwin, IBM,
David Bacon, IBM,
David Grove, IBM,
Chet Murthy, IBM,

While Java provides many software engineering benefits, it lacks a coherent module system and instead provides only packages (which are primarily a name space mechanism) and classloaders (which are very low-level). As a result, large Java applications suffer from unexpected interactions between independent components, require complex CLASSPATH definitions, and are often extremely complex to install and maintain.

We have implemented a module system for Java that is implemented with class loaders but provides a much higher-level interface. High-level properties can be specified in a module definition and are enforced by the module system as new modules are loaded.

To experimentally validate the ability of ModJava to properly handle the complex module inter-relationships found in large Java server systems, we replaced the classloader mechanisms of Apache Tomcat 4.1.18 with 30 ModJava modules. The modified Tomcat is functionally identical to the original, but requires no CLASSPATH definitions, and will operate correctly even if user code loads a different version of a module used by Tomcat, such as the Xerces XML parser. Furthermore, by making a small change to the Java core libraries enabled by ModJava, we obtained a 30% performance improvement in a servlet microbenchmark.

Model Driven Architecture: How far have we come, how far can we go?

Wednesday, 29 October – 15:30-17:00

Granville Miller (Chair), Borland,
Andy Evans, Xactium Limited,
Ivar Jacobson, JacZone,
Henrik Jondel, Borland,
Allan Kennedy, Kennedy Carter,
Stephen Mellor, Project Technology,
Dave Thomas, Bedarra Research Labs,

Model Driven Architecture (MDA) is a technology that has been in the process of evolution for many years. Today, many vendors are now producing products that support MDA. We are hearing more and more success stories that indicate that this technology is the "real deal". But, with the failed promises of CASE in the late 1980's, many people still have questions about how much of an application can be generated from models and constraint languages. Is MDA really capable of generating enterprise applications? What are the technologies are available to implement MDA? Here is your opportunity to ask the experts the questions that are necessary to convince you of the validity of this new technology. Each of these panelists has been intricately involved in building the underlying foundations of Model Driven Architecture and its implementation.

Real-World Objects

Wednesday, 29 October – 15:30-17:00

15:30 - 16:00
Programming with Non-Heap Memory in RTSJ

Greg Bollella, Sun Microsystems Laboratories
Tim Canham, Jet Propulsion Laboratory, California Institute of Technology
Vanessa Carson, Jet Propulsion Laboratory, California Institute of Technology
Virgil Champlin, Jet Propulsion Laboratory, California Institute of Technology
Daniel Dvorak, Jet Propulsion Laboratory, California Institute of Technology
Brian Giovannoni, School of Computer Science, Carnegie Mellon University
Kenny Meyer, Jet Propulsion Laboratory, California Institute of Technology
Alex Murray, Jet Propulsion Laboratory, California Institute of Technology
Kirk Reinholtz, Jet Propulsion Laboratory, California Institute of Technology

The Real-Time Specification for Java (RTSJ) provides facilities for deterministic, real-time execution in a language that is otherwise subject to variable latencies in memory allocation and garbage collection. In particular, RTSJ introduces two kinds of non-heap memory area (ScopedMemory and ImmortalMemory) and a kind of real-time thread (NoHeapRealtimeThread) that, together, can satisfy very demanding timeliness requirements because they run without interference from the garbage collector. What new RTSJ programmers quickly discover, though, is that the RTSJ "assignment rules"—governing references from one kind of memory to another—force designers to think more carefully about how data flows among the components of a system. In particular, for the hard real-time portions of a system, the normal Java practice of freely passing around references and letting the garbage collector recognize when an object is no longer needed simply doesn't apply.

This report discusses two related design issues that all RTSJ users will face: how to explicitly manage non-heap memory and how to communicate information among components without violating RTSJ’s assignment rules. These issues are addressed from a practitioner’s viewpoint in the context of designing real-time control loops for driving and steering a 6-wheel Mars rover. The solution described here is termed “scoped memory scratchpads” and is believed to provide a good combination of Java-natural style and safety from programmer error. In this approach, transient data is allocated within a scoped memory area that is emptied later when all Runnables exeunt the memory area. Data that must be preserved across cyclic executions use memory pools, restricted so that all pool management is confined to a single component and is thus much easier to design and verify. Importantly, by maintaining a distinction between application-specific software and framework software, some of the details and discipline of using RTSJ can be hidden from application programmers.

16:00 - 16:30
The Parks PDA: A Handheld Device for Theme Park Guests in Squeak

Yoshiki Ohshima, Twin Sun, Inc.,
John Maloney, MIT Media Lab,
Andy Ogden, Strategy, Design, and Development Consulting,

The Parks PDA is a lightweight, handheld device for theme park guests that functions as a combination guidebook, map, and digital camera. Together with a small team of artists and designers, we created a prototype Parks PDA and content for a three hour guest experience, including a camera interface, a hyper-linked guide book, three games, an animal spotters guide, a cross-referenced map, animated movies with lip-synched sound, a ride reservation system, and more. Over 800 visitors to Disney's Animal Kingdom theme park tested the Parks PDA over a two week period.

Developing the software for this test posed a number of challenges. The processor and memory of the target device were slow, the screen was small, and we had only three months of development time.

We attacked these problems using Squeak, a highly-portable, open source Smalltalk implementation. We ported Squeak to the target device and used it to provide nearly bit-identical behavior across four different platforms. This supported a cross-platform development style that streamlined the production of both software and content. We created a tiny user interface and application framework for pen-based devices and implemented a simple card-stack media editor and player using it. We isolated and fixed several performance problems.

The project was completed on time and guest response was favorable. Looking back, we can identify ten aspects of Squeak that contributed to the success of the project. While we feel that Squeak was the ideal tool for this job, many of the key benefits of Squeak could be enjoyed using other object-oriented languages such as Java J2ME or Python, especially if the source code of the virtual machine is available.

The Biology of Information

Wednesday, 29 October – 15:30-17:00

Walter Fontana, Santa Fe Institute,

Walter Fontana is a research professor in residence at the Santa Fe Institute, has served as a Member of the Institute for Advanced Study in Princeton, NJ, with the Program in Theoretical Biology, and was a member of the University of Vienna faculty at the Institute for Theoretical Chemistry and Molecular Structural Biology.

OOPSLA Keynote - The Internet Paradigm Shift

Thursday, 30 October – 8:30-10:00

Tim O'Reilly, O'Reilly and Associates,

The computer industry has gone through a sea change in the past few years. The killer applications of the web era turned out not to be PC-based software packages like the web browser, but web hosted applications like google, mapquest and These applications are built on top of Linux and Apache, yet they are themselves fiercely proprietary. But what would most developers do with their source code? These massive systems are valuable for their data as much as for their programs. And by opening up XML web services APIs to that data, the most innovative of these sites are creating new opportunities for hackers to re-use that data and "scratch their own itch." What's more, as constantly updated services, these applications operate on very different timelines and processes than conventional software development. One of the greatest challenges for developers in the next few years is to understand and adapt to the paradigm shift implicit in network computing, and to shed the legacy thinking of the desktop era.


Tim O'Reilly is founder and president of O'Reilly & Associates, thought by many to be the best computer book publisher in the world. In addition to publishing pioneering books like Ed Krol's The Whole Internet User's Guide & Catalog (selected by the New York Public Library as one of the most significant books of the twentieth century), O'Reilly has also been a pioneer in the popularization of the Internet. O'Reilly's Global Network Navigator site (GNN, which was sold to America Online in September 1995) was the first Web portal and the first true commercial site on the World Wide Web.

Tim has been an activist for internet standards and for Open Source software. He has led successful public relations campaigns on behalf of key internet technologies, helping to block Microsoft's 1996 limits on TCP/IP in NT Workstation, organizing the "summit" of key free software leaders where the term "Open Source" was first widely agreed upon, and, most recently, organizing a series of protests against frivolous software patents. Tim received Infoworld's Industry Achievement Award in 1998 for his advocacy on behalf of the Open Source community.

Garbage Collection 1

Thursday, 30 October – 8:30-10:00

8:30 - 9:00
Mostly Concurrent Garbage Collection Revisited

Katherine Barabash, IBM Haifa Research Laboratory,
Yoav Ossia, IBM Haifa Research Laboratory,
Erez Petrank, Technion - Israel Institute of Technology,

The mostly concurrent garbage collection was presented in the seminal paper of Boehm et al. With the deployment of Java as a portable, secure and concurrent programming language, the mostly concurrent garbage collector turned out to be an excellent solution for Java's garbage collection task. The use of this collector is reported for several modern production Java Virtual Machines, and it has been investigated further in academia.

In this paper, we present a modification of the mostly concurrent collector which improves the throughput, the memory footprint, and the cache behavior of the collector without foiling the other good qualities (such as short pauses and high scalability). We implemented our solution on the IBM production JVM and obtained a performance improvement of up to 22.8%, a reduction in the heap consumption by up to 13.6%, and no substantial change in the (short) pause times. The modified algorithm has subsequently been incorporated into the IBM production JVM.

9:00 - 9:30
An On-the-Fly Mark and Sweep Garbage Collector Based on Sliding Views

Hezi Azatchi, Technion - Israel Institute of Technology,
Yossi Levanoni, Technion - Israel Institute of Technology,
Harel Paz, Technion - Israel Institute of Technology,
Erez Petrank, Technion - Israel Institute of Technology,

With concurrent and garbage collected languages like Java and C# becoming popular, the need for a suitable non-intrusive, efficient, and concurrent multiprocessor garbage collector has become acute. We propose a novel mark and sweep on-the-fly algorithm based on the sliding views mechanism of Levanoni and Petrank. We have implemented our collector on the Jikes Java Virtual Machine running on a Netfinity multiprocessor and compared it to the concurrent algorithm and to the stop-the-world collector supplied with Jikes JVM. The maximum pause time that we measured with our benchmarks over all runs was 2ms. In all runs, the pause times were smaller than those of the stop-the-world collector by two orders of magnitude and they were also always shorter than the pauses of the Jikes concurrent collector. Throughput measurements of the new garbage collector show that it outperforms the Jikes concurrent collector by up to 60%. As expected, the stop-the-world does better than the on-the-fly collectors with results showing about 10% difference.

On top of being an effective mark and sweep on-the-fly collector standing on its own, our collector may also be used as a backup collector (collecting cyclic data structures) for the Levanoni-Petrank reference counting collector. These two algorithms perfectly fit sharing the same allocator, a similar data structure, and a similar JVM interface.

9:30 - 10:00
Heap Compression for Memory-Constrained Java Environment

Guangyu Chen, The Pennsylvania State University,
Mahmut Kandemir, The Pennsylvania State University,
Naraya Vijaykrishnan, The Pennsylvania State University,
Janie Irwin, The Pennsylvania State University,
Bernd Mathiske, Sun Microsystems, Inc.,
Mario Wolczko, Sun Microsystems, Inc.,

Java is becoming the main software platform for consumer and embedded devices such as mobile phones, PDAs, TV set-top boxes, and in-vehicle systems. Since many of these systems are memory constrained, it is extremely important to keep the memory footprint of Java applications under control.

The goal of this work is to enable the execution of Java applications using a smaller heap footprint than that possible using current embedded JVMs. We propose a set of memory management strategies to reduce heap footprint of embedded Java applications that execute under severe memory constraints. Our first contribution is a new garbage collector, referred to as the Mark-Compact-Compress (MCC) collector, that allows an application to run with a heap smaller than its footprint. An important characteristic of this collector is that it compresses objects when heap compaction is not sufficient for creating space for the current allocation request. In addition to employing compression, we also consider a heap management strategy and associated garbage collector, called MCL (Mark-Compact-Lazy Allocate), based on lazy allocation of object portions. This new collector operates like the conventional Mark-Compact (MC) collector, but takes advantage of the observation that many Java applications create large objects, of which only a small portion is actually used. In addition, we also combine MCC and MCL, and present MCCL (Mark-Compact-Compress-Lazy Allocate), which outperforms both MCC and MCL.

We have implemented these collectors using KVM, and performed extensive experiments using a set of ten embedded Java applications. We have found our new garbage collection strategies to be useful in two main aspects. First, they reduce the minimum heap size necessary to execute an application without out-of-memory exception. Second, our strategies reduce the heap occupancy. That is, at a given time, they reduce the heap memory requirement of the application being executed. We have also conducted experiments with a more aggressive object compression strategy and discussed its main advantages.

Agile Management—An Oxymoron?

Thursday, 30 October – 8:30-10:00

Lougie Anderson (Chair), Sabrix, Inc.,
Glen Alleman, CH2M Hill,
Kent Beck, Three Rivers Institute,
Joe Blotner, Sabrix, Inc.,
Ward Cunningham, Cunningham & Cunningham,
Mary Poppendieck, Poppendieck, LLC,
Rebecca Wirfs-Brock, Wirfs-Brock Associates,

"Self-directed team" is one of the mantras of Agile Methodologies. Self-direction means that the team's manager is relegated to a facilitator role with little or no influence over day-to-day activities. For example, Kent Beck has written that the manager of an XP project can do four things: ask for estimates on cost and results, move people around among projects, ask for status reports, and cancel the project. Agile literature in general says that managers shouldn't be directly involved in analysis, design, coding, testing or integration. They may (but only occasionally!) facilitate the process between the customer and the developers; and it would be nice if they provided food and toys to keep the team happy. It appears, then, that the agile manger is expected to hover on the fringes of a project asking a few questions and throwing in goodies—but with ultimate power (cancellation) in her hip pocket. This scenario makes one wonder. Do managers really matter to the success of an agile project? Are they superfluous? What happens when managers step over the prescribed line—does it mean that the end of Agile Methodology as we know it and as handed down by the Agile Manifesto? The panel will explore this ticklish terrain by answering the following questions: Why Agile Methods and managers don't mix. Or do they? What can/should managers do in an agile environment? Under what conditions are managers an absolute requirement in an agile environment? (e.g. Government applications?) Do good management techniques apply to both Agile and non-Agile environments? Is management a dead-end profession in an Agile world?

Garbage Collection 2

Thursday, 30 October – 10:30-12:00

10:30 - 11:00
MarkCopy: Fast Copying GC with Less Space Overhead

Narendran Sachindran, University of Massachusetts Amherst,
Eliot Moss, University of Massachusetts Amherst,

Copying garbage collectors have several advantages over non-copying collectors, including cheap allocation and prevention of fragmentation. However, in order to provide completeness (the guarantee to reclaim each garbage object eventually), standard copying collectors require space equal to twice the size of the maximum live data for a program.

We present a mark-copy collection algorithm that extends generational copying collection and significantly reduces the heap space required to run a program. The new collector runs in space that is nearly a factor of two smaller than that required by standard copying garbage collectors, increasing the range of applications that can use copying garbage collection. We show that when the collector is given the same amount of space as a generational copying collector, it improves program execution time of Java benchmarks by 20-85% in tight heaps and by 5-10% in moderate size heaps.

We also compare the performance of the mark-copy collector with that of a mark-sweep collector. We find that, for several benchmarks, the mark-copy collector can run in heaps comparable in size to the minimum heap space required by the mark-sweep collector. We also find that for some benchmarks the mark-copy collector is significantly faster than a mark-sweep collector in tight heaps.

11:00 - 11:30
Ulterior Reference Counting: Fast Garbage Collection without a Long Wait

Steve Blackburn, Australian National University,
Kathryn McKinley, University of Texas at Austin,

General purpose garbage collectors have yet to combine short pause times with fast throughput. For example, non-concurrent generational collectors can achieve high throughput and have modest average pause times, but occasionally collect the whole heap and consequently incur long pauses. On the other extreme, concurrent collectors, including reference counters, attain short pause times with significant performance penalties. This paper introduces GenRC, which combines copying generational collection and reference counting the older generation to achieve both goals in a uniprocessor setting. Key to our algorithm is a generalization of deferred reference counting which allows GenRC to safely ignore mutations to nursery objects. GenRC thus restricts copying and reference counting to the object demographics for which they perform well. The contiguous allocation of a copying collector is extremely fast, and collection time is proportional to the live objects whose survival rates are usually low in a fixed size nursery. Reference counting time is proportional to object mutations and the number of dead objects, both of which are typically low for older objects. We further restrict the time spent reference counting with collection triggers and buffering. We compare GenRC using a fixed nursery (FG-RC) with pure reference counting (RC), high performance copying and mark-sweep fixed nursery generational (FG-MS), and pure mark-sweep collectors (MS). We show that FG-RC is competitive with FG-MS on throughput, and attains much lower average and maximum pause times.

11:30 - 12:00
Connectivity-Based Garbage Collection

Martin Hirzel, University of Colorado at Boulder,
Amer Diwan, University of Colorado at Boulder,
Matthew Hertz, University of Massachusetts Amherst,

We introduce a new family of connectivity-based garbage collectors (CBGC) that are based on potential object-connectivity properties. The key feature of these collectors is that the placement of objects into partitions is determined by performing one of several forms of connectivity analyses on the program. This enables partial garbage collections, as in generational collectors, but without the need for any write barrier.

The contributions of this paper are 1) a novel family of garbage collection algorithms based on object connectivity; 2) a detailed description of an instance of this family; and 3) an empirical evaluation of CBGC using simulations. Using simulations allows us to explore a broad range of possibilities for CBGC, ranging from simplistic ones that determine connectivity based on type information to oracular ones that use run-time information to determine connectivity. Our experiments with the oracular CBGC configurations give us an indication of the potential for CBGC and also identify weaknesses in the realistic configurations. We found that even the simplistic implementations beat state-of-the-art generational collectors with respect to some metrics (e.g., pause times and memory footprint).

Object-Oriented Success Stories: "Learning from our Failures"

Thursday, 30 October – 10:30-12:00

Joseph Yoder (Chair), The Refactory Inc. & The University of Illinois,
Ralph Johnson, The Refactory Inc. & The University of Illinois,
Steven Wingo, Southern Company,
Ron Jeffries,,
Linda Rising, Independent Consultant,

Beneath the buzz around methodologies, languages and technologies, the last eighteen years at OOPSLA have seen countless object-oriented success stories, large and small. This fishbowl will provide OOPSLA attendees to bear witness to these victories, and tell these tales at last. Similarly we propose a follow-up fishbowl that discusses our failures. Just as much (if not more) can be learned from failures as can from successes.

Panel: Reuse Repositories and Reuse—The Realities

Thursday, 30 October – 10:30-12:00

This session starts with a photo-essay on components, reuse, value, and beauty, and concludes with a panel on libraries, repositories, and reuse.

Without a Name: A Reusable Component Repository

Robert Biddle, Victoria University of Wellington, New Zealand,
Angela Martin, Victoria University of Wellington, New Zealand,
James Noble, Victoria University of Wellington, New Zealand,

The Story - this essay concerns components, reuse, value, and beauty. All photographs were taken at No Name Building Recyclers, Wellington, New Zealand, with the kind permission of the owners.

Panel: Reuse Repositories and Reuse—The Realities

Dave Thomas, Bedarra Research Labs & Carleton University,
Brian Barry, Bedarra Corp &,
Ivar Jacobson, JacZone,
Linda Northrop, Software Engineering Institute,
Clemens Szyperski, Microsoft Research,
& others

This panel (part of the 2003 Onward! program) will discuss repositories, and reuse. While there is so much hype and noise about components and model repositories, metadata and reuse there is still very little understood about how hard it is to design for reuse and to encourage systematic reuse both in terms of culture, design, and supporting tools and technology. Like standards we have a plethora of component models to choose from and yet the long predicted component market place has yet to occur or has it already happened?

Our educational colleagues outside of the our community are blindly following our rhetoric to build eLearning repositories so Learning Objects can be snapped together just like the OO folks do it! Similarly object zealots are arguing for OO XML, OO CSS to facilitate reuse OO style.

Our panelists have participated in both the euphoric vision as well as the pragmatic realities of repositories reuse in practice. The panel has been asked to discuss their future vision for reuse and repositories and in particular what key social, business and technical enablers will facilitate significant reuse or render the matter irrelevant.


Thursday, 30 October – 10:30-11:30

10:30 - 11:00
Declaring and Checking Non-null Types in an Object-Oriented Language

Manuel Fähndrich, Microsoft Research,
Rustan Leino, Microsoft Research,

Distinguishing non-null references from possibly null references at the type level can detect null-related errors in object-oriented programs at compile-time. This paper gives a proposal for retrofitting a language such as C# or Java with non-null types. It addresses the central complications that arise in constructors, where declared non-null fields may not yet have been initialized, but the partially constructed object is already accessible. The paper reports experience with an implementation for annotating and checking null-related properties in C# programs.

11:00 - 11:30
Object Equality Profiling

Darko Marinov, MIT Laboratory for Computer Science,
Robert O'Callahan, IBM T.J. Watson Research Center,

We present Object Equality Profiling (OEP), a new technique for helping programmers discover optimization opportunities in programs. OEP discovers opportunities for replacing a set of equivalent object instances with a single representative object. Such a set represents an opportunity for automatically or manually applying optimizations such as hash consing, heap compression, lazy allocation, object caching, invariant hoisting, and more. To evaluate OEP, we implemented a tool to help programmers reduce the memory usage of Java programs. Our tool performs a dynamic analysis that records all the objects created during a particular program run. The tool partitions the objects into equivalence classes, and uses collected timing information to determine when elements of an equivalence class could have been safely collapsed into a single representative object without affecting the behavior of that program run. We report the results of applying this tool to benchmarks, including two widely used Web application servers. Many benchmarks exhibit significant amounts of object equivalence, and in most benchmarks our profiler identifies optimization opportunities clustered around a small number of allocation sites. We present a case study of using our profiler to find simple manual optimizations that reduce the average space used by live objects in two SpecJVM benchmarks by 47% and 38% respectively.

Eclipse and The Dark Side of the Moon

Thursday, 30 October – 13:30-15:00

Erich Gamma, IBM

Eclipse ( is an open source universal tool platform for anything and nothing in particular. This talk looks at Eclipse inside the corona.


Erich Gamma leads the Eclipse Java Development tools project and is a member of the Eclipse project management committee. He is also a member of the Gang of Four, which is known for their book: Design Patterns - Elements of Reusable Object-Oriented Software. Erich has paired with Kent Beck to develop JUnit, a popular testing tool for Java. Before joining OTI he was working at Taligent on a never shipped C++ development environment. Erich started with object-oriented programming over 20 years ago as a the co-author of ET++ one of the first large scale C++ application frameworks.

Erich Gamma is the site lead of the IBM OTI Lab in Zurich, Switzerland. He has a doctorate in computer science from the University of Zurich. Erich was a Swiss ski instructor but never a fan of Pink Floyd.

Transactions and Persistence

Thursday, 30 October – 13:30-15:00

13:30 - 14:00
Saving the World from Bad Beans: Deployment-Time Confinement Checking

Dave Clarke, Utrecht University,
Michael Richmond, IBM Almaden Research Center,
James Noble, Victoria University of Wellington,

The Enterprise JavaBeans (EJB) framework requires developers to preserve architectural integrity constraints when writing EJB components. Breaking these constraints allows components to violate the transaction protocol, bypass security mechanisms, disable object persistence, and be susceptible to malicious attacks from other EJBs. We present an object confinement discipline that allows static verification of component integrity as they are deployed into an EJB server. The confinement rules are simple for developers to understand, require no annotation to the code of EJB components, and enforcement of these rules can be incorporated efficiently into existing EJB servers.

14:00 - 14:30
Language Support for Lightweight Transactions

Timothy Harris, University of Cambridge Computer Laboratory,
Keir Fraser, University of Cambridge Computer Laboratory,

Concurrent programming is notoriously difficult. Current abstractions are intricate to use and make it hard to design computer systems that are reliable and scalable. We argue in favour of a practical declarative style of concurrency control in which programmers directly indicate the safety properties that they require. In essence, the programmer demarks sections of code which execute within lightweight software-based transactions which commit atomically and exactly once.

These transactions can update shared data, instantiate objects, invoke library features and so on. They can also block, waiting for arbitrary boolean conditions to become true. These features support general-purpose shared data structures such as hashtables, queues and buffers. In general, no performance penalty is incurred for memory accesses outside transactions. Furthermore, transactions which do not access the same shared memory locations can commit concurrently.

We present a detailed design of this proposal along with an implementation and evaluation. We argue that the resulting system (i) is easier for mainstream programmers to use, (ii) prevents lock-based priority-inversion and deadlock problems and (iii) offers compelling performance.

14:30 - 15:00
Lazy Modular Upgrades in Persistent Object Stores

Chandrasekhar Boyapati, MIT Laboratory for Computer Science,
Barbara Liskov, MIT Laboratory for Computer Science,
Liuba Shrira, MIT Laboratory for Computer Science,
Chuang-Hue Moh, MIT Laboratory for Computer Science,
Steve Richman, MIT Laboratory for Computer Science,

Persistent object stores require a way to automatically upgrade persistent objects. Automatic upgrades are a challenge for such systems. Upgrades must be performed in a way that is efficient both in space and time, and that does not stop application access to the store. In addition, however, the approach must be modular: it must allow programmers to reason locally about the correctness of their upgrades similar to the way they would reason about regular code. This paper provides solutions to both problems.

The paper first defines upgrade modularity conditions that any upgrade system should satisfy to support local reasoning about upgrades. The paper then describes a new approach for executing upgrades efficiently while satisfying the upgrade modularity conditions. The approach exploits object encapsulation properties in a novel way. The paper also describes a prototype implementation and shows that our upgrade system imposes only a small overhead on application performance.

What's so eXtreme About Doing Things Right?

Thursday, 30 October – 13:30-15:00

Steve Berczuk (Chair), Independent Consultant,
Neil Harrison, Avaya Labs,
Kevlin Henney, Curbralan,
Joshua Kerievsky,, Industrial Logic,
Linda Rising, Independent Consultant,
Ken Schwaber, ADM,
Bobby Woolf, Independent Consultant,

Agile Methods are advocated as a way of producing better software. Advocates of agile methods suggest that practices such as keeping in close communication with your customers, frequent integration, and frequent assessment of project status will enable us to produce software that has value for the customer—quality software. It's hard to argue with that. But why is this any different than simply "good" software development practice? Why does saying "Scrum" "Agile" or "XP" grab peoples' attention? Why does it take a name for useful practices to be accepted? This panel will help us understand the role of hype in getting useful practices accepted or rejected. We will explore why it is that these good ideas have not been more widely used. Some of the questions that the panel and the audience will explore are: Why do we ignore proven practices until we see them packaged as a "method?" Can we do something different in the workplace or in school to teach these practices? Or is it the case that these practices are not universally good? This panel talks about agility in a different context than what is typical: we won't just discuss what agile practices are. We will explore why they are not more widely adopted, especially when not packaged as part of a "named" method like XP, and we will discuss why projects suffer even when the methods that can help them are well known. This panel will provide an entertaining and thought provoking forum for discussing an issue that is ever present in the world of technology: the role of hype. We concentrate on agile practices, moving beyond simply enumerating them, to discussing why they are not more widely adopted.

Application Servers: One Size Fits All ... Not?

Thursday, 30 October – 13:30-15:00

Gail E. Harris (Chair), Instantiated Software Inc.,
Jeromy Carrière, Microsoft Corporation,
John Crupi, Sun Microsystems,
David Leibs, Oracle Corporation,
Fred Nagy, Solutions In Context,
Martin Nally, IBM Corporation,

In the beginning there was machine language, followed by assembly language, formula translation, and eventually procedural programming, to organize the chaos. And then objects were introduced, to hide information. Soon Client/Server and multi-tier applications were conceived to separate data concerns from business logic concerns and user interface concerns. Later, these objects were distributed geographically to optimize hardware resources. And now, we have application servers, to simplify scaling up a system for large volumes, improved response times, impeccable reliability, and high availability. Application servers house the business logic, operating on data from a different server, and responding to requests from any source. But these Application Servers come in all shapes, flavors, and sizes. What is a developer to do? This panel will explore issues comparing application server technologies and questions about their appropriate use in different contexts.