ACCU Home page ACCU Conference Page
Search Contact us ACCU at Flickr ACCU at GitHib ACCU at Facebook ACCU at Linked-in ACCU at Twitter Skip Navigation

ACCU 2003 Speakers

Andrei Alexandrescu

Andrei is a world-class expert in software development using C++. In the C++ community, he is best known for his book, Modern C++ Design (Addison Wesley, 2001). Also, Andrei is a former columnist for the C++ Report, a columnist for C/C++ Users Journal, and a sought-after speaker at conferences worldwide.
After working in large-scale projects ranging from financial software on Wall Street to networking software to user interfaces, Andrei is pursuing a Ph.D. in Computer Science at University of Washington.

Prying Eyes: Generic Observer Implementations in C++

The Observer design pattern has many ways of being applied and very different implementation favoring speed, information flow, control flow, flexibility. This talk categorizes Observer implementations and builds a policy-based implementation for generic Observers. (It also offers a sneak preview into Andrei's upcoming book, The Return of C++.)

Honey, I Shrunk the Threads: Compile-time checked multithreaded transactions in C++

Would you like to have your compiler tell you "Error: This call will lead to deadlock"? This talk presents new idioms for C++ multithreaded programs, including multithreaded transactions involving multiple objects. You can have your compiler detect not only race conditions, but deadlocks as well.


David Abrahams

David Abrahams is a founding member and moderator of Boost, and an active member of the wider open-source community. He has been an ANSI/ISO C++ committee member since 1996, when he developed a workable theory of exception-safety and an exception-safety specification for the C++ standard library. In his 14-year career he has developed applications for the desktop and embedded devices in the fields of music software, speech recognition, and circuit simulation. In 2001 He founded Boost Consulting, a company dedicated to providing professional support and development services for the Boost C++ libraries and associated tools.

Binding C++ to Python with the Boost Python Library

In many ways, the Python language makes a perfect complement to C++: as an interpreted, dynamically-typed language it allows the programmer to quickly and compactly build large-scale applications that apply idioms from C++ generic programming. C++ and Python can be combined in several popular ways: for example C++ can be used to provide high-performance core components for Python applications, or Python can be used as a plug-in or extension language for applications written in C++. Effectively binding Python and C++ requires a deep understanding of the internals and idioms of both languages. The Boost Python library is a framework of tools which allow users to quickly expose C++ types and functions to Python by writing a kind of simple "Interface Definition Language" (IDL) directly in C++ code. By using the compile-time capabilities of C++ and the techniques of metaprogramming, Boost.Python avoids many of the problems of other wrapping systems, which must implement most of the internals of a C++ compiler to work automatically. The library also includes facilities for manipulating Python objects from C++ using a "Python-like" interface.

Metaprogramming and the Boost Metaprogramming Library

The Boost C++ template metaprogramming library (MPL) is an extensible compile-time framework of algorithms, sequences and metafunction classes. The library brings together important abstractions from the generic and functional programming worlds to build a powerful and easy-to-use toolset which makes template metaprogramming practical for real-world environments. The MPL is heavily influenced by its run-time equivalent – the Standard Template Library (STL), a part of the C++ standard library. Like the STL, it defines an open conceptual and implementation framework which can serve as a foundation for future contributions in the domain. The library's fundamental concepts and idioms enable the user to focus on solutions without navigating the universe of possible ad-hoc approaches to a given metaprogramming problem, even if no actual MPL code is used. It also provides a compile-time lambda expression facility enabling arbitrary currying and composition of class templates, a feature whose runtime counterpart is often cited as missing from the STL. This talk explains the motivation, usage, design, and implementation of the MPL with examples of its real-life applications, and offers some lessons learned about C++ template metaprogramming.


Willem Wakker

Willem Wakker has been active in the world of formal standards and programming language standardisation since 1988. He also has extensive industrial experience in this area having joined ACE in 1977 and led their compiler development team until 1995 when he was appointed Director of ACE's Consulting operation. Willem is currently in WG14 and is the project editor of the Embedded C Technical Report. It should be noted that ACE defined and introduced the predecessor of Embedded C, namely DSP-C.

The Embedded C Extensions to C

Embedded C is the name of a language extension to currently being formalized in a technical report by the ISO. Embedded C originates from an industry initiative called DSP-C developed by ACE. It aims to provide C application programmers with access to common performance increasing features of processors used in the domain of DSP and embedded processing.

Embedded C adds fixed-point data types, memory qualifiers and hardware I/O to C. Fixed point data types are frequently used in embedded applications and routinely present in DSP processors. Embedded C adds this primitive type to allow the compiler to make the connection between the two. Similarly, multiple memory banks, typically present in embedded processors to provide the ALU with enough bandwidth, can be addressed using memory qualifiers. Hardware I/O provides a standardized abstraction layer for accessing hardware ports while still allowing for maximal efficiency. This improves portability of, for example, device driver code.

Embedded C is currently being developed by the ISO C working group (WG 14) as a technical report. It is not part of the C language yet. Its aim is to provide common ground for initiatives that currently exist within the industry to give the programmer control over certain hardware features. The report is an important step in streamlining these initiatives and making the development of application support for these features more efficient. The technical report is expected to be ratified in 2003.

The presentation will explain about the meaning and use of the new language extensions, including examples of their use and resulting DSP processor code. Additionally, the rationale behind the design will be discussed, including alternative proposals that did not make it, and the relation between Embedded C and C++.

The presentation will discuss the meaning and use of the language extensions and provide examples of each. Additionally, the motivation behind Embedded C and the rationale of certain design decisions will be explained as well as the status of the technical report. The relation with C++ is also addressed.


Beman Dawes

Beman is the founder of Boost (, the source of portable free  C++ libraries and a proving ground for future C++ Standard Library additions. He has been a voting member of the ANSI/ISO C++ Standards Committee since 1992, and chaired the Library Working Group for five years.  He develops portable C++ libraries for a living, and is the author of the StreetQuick geographic atlas library.

Multi-Platform Software Development; Lessons from the Boost libraries

Developing software in one platform or environment (operating system, compiler, company, department, whatever) and then later trying to port it  to some other platform is often a recipe for disaster. Yet experience both at Boost and with industrial applications shows that certain software development practices can remove much of the risk from cross-platform development. Learn about these practices and how to apply them to real-world software development.

Although based on experience with C++ cross-platform efforts, much of the talk focuses on development practices which apply to any programming language.

Before the conference, please send a description of your cross-platform problem to <> so solutions can be discussed during the talk.


Frank Buschmann

Frank Buschmann is senior principal engineer at Siemens Corporate Technology in Munich, Germany. His interests include Object Technology, Frameworks and Patterns. Frank has been involved in many software development projects. He is leading Siemens' pattern research activities. Frank is co-author of   "Pattern-Oriented Software Architecture – A System of  Patterns" and "Pattern-Oriented Software Architecture – Patterns for Concurrent and Networked Objects"

Pattern Writing: Live and Direct

(co-presented with Kevlin Henney)

The overwhelming majority of developers familiar with patterns would consider themselves consumers rather than producers of patterns literature. How would you get started writing a pattern? And how would you continue? One way to understand something is to take it apart; another is to make one. This tutorial aims to do both of these for patterns.
It starts with an anatomical dissection of the common pattern, and places patterns in the context of a pattern language. The two presenters then pair-write patterns in an actual pattern language, switching between writing, making meta-level comments about the issues encountered, and taking suggestions and questions from the attendees.


Greg Colvin

I Dr. Colvin learned to program in 1972, on a PDP-8. He has been coding C and doing object oriented design since 1984, C++ and generic design since 1988, has served on WG21 since 1990, is a founding Boost member, and has recently joined WG 14. He has worked in the Oracle JVM group since 2000, and is Oracle's C/C++ liaison for ISO. He is also a dedicated blues musician, recording engineer, Tennessee Walking Horse breeder, and antiwar activist in what little spare time remains.

In the Spirit of C [Keynote]

C++ and Java owe much to their C heritage, and the usefulness of C has only been increased by the success of its descendants.  These languages serve a community of related interests, and in my opinion need both constraints on compatibility and room to grow in order to continue to be of service.  The most important maxim of the Spirit of C is "Trust the Programmer."  That maxim imposes a corollary obligation on us as designers and standardizers to provide the programmer with trustworthy and effective tools for the job at hand.  We will explore where and how these languages are effective today, and what we need to do to maintain that effectiveness into the future.


Daveed Vandevoorde

David Vandevoorde is an engineer at the Edison Design Group, where he codeveloped the first 100% ISO-compliant C++ compiler. He is an active member of the ANSI C++ Standards Committee, and a cofounder of the newsgroup comp.lang.c++.moderated. He is the author (with Nicolai M. Josuttis) of "C++ Templates: The Complete Guide". A graduate of the Brussels Free University and the Rensselaer Polytechnic Institute, his interests include algorithm development, programming languages, and teaching. See 


Reflective Metaprogramming

We're nearing one decade of template metaprogramming in C++. In that decade several metaprogramming challenges have been resolved, several others have been found unsolvable using template-based techniques, but perhaps most of all, metaprogramming has been found to be a sound tool for program development. Furthermore, reflection appears to be a very desirable facility in the metaprogramming arsenal. In this talk I will sketch a complete reflective metaprogramming language extension for C++. The ideas underlying this extension draw from C++ template metaprogramming experience and from a personal research language called Xroma.


Jon Jagger

Hi, I'm Jon Jagger, an independent software consultant/trainer/mentor specialising in C#, C++, Java, OO, design, patterns, and process improvement. I am a UK C#, C++, and C standards panel member and a regular contributor to the ACCU Overload journal. My interests include training excellence, design, problem solving, and monty python (required knowledge for all software developers). I'm very very good at sleeping, breathing, and drinking. All of which I practice a lot.


I wrote most of a 5 day instructor led training course on C# that now forms part of the official Microsoft curriculum (Introduction to C# Programming, course 2124).

I co-authored (with John Sharp) the Microsoft Press book Visual C#.NET Step by Step which is getting good reviews on Amazon.

I converted the ECMA C# language specification into a hyperlinked HTML presentation using PERL, XML, and XSL (available from my website).

Sauce: An OO recursive descent parser; its design and implementation.

This talk recounts the design, patterns in, and implementation of Sauce. Sauce is an extensible tool that you can use to detect syntactic and lexical fluff in your code. (Sauce came into being when one day I imagined a tool that could check all your binary operators were braced by matching whitespace. All the lint tools I could find were purely syntactic!) I will also discuss the differences encountered, and lessons learned, when implementing Sauce in more than one language (C++ and C#).   


Nico Josuttis

Nicolai Josuttis ( is an independent systems architect, author, and consultant. He designs mid-sized and large software systems for the telecommunication, traffic, finance, and manufacturing industries. He is well known both in the C++ Community and to attendees at ACCU Conferences. He not only speaks and writes with authority about C++ (being the author of 'The C++ Standard Library') but is also an innovative presenter.
He has also written other books and articles about object-oriented software development and programming in general. He is a partner at System Bauhaus, a German group of recognized object-oriented system development experts

Secrets and Pitfalls of Templates

(co-presented with David Vandevoorde)
Although templates have been part of C++ for well over a decade (and available in various forms for almost as long), they still lead to misunderstanding, misuse and/or controversy. However, the effective C++ programmer should know about some important C++ template issues that are not widely known. This talk presents some of them.

Plenty People Programming: C++ Programming in a Group

A session, where the audience writes some little programs (using me as the "typing system") for tasks provided by me and where we try out different compilers. Thus, I give a task and wait for input from the audience to put it into the laptop.
And we see, what happens when we run different compilers (such as EDG, G++, Visual C++).


Jutta Eckstein

Jutta Eckstein (, is an independent consultant and trainer from Munich, Germany. Her know-how in agile processes is based on over ten years experience in developing object-oriented applications. She is an experienced XP coach and trainer.  She worked with teams of different sizes mainly in the finance industry to help them using agile processes successfully. Besides engineering software she has been designing and teaching OT courses in industry.  Having completed a course of teacher training and led many 'train the trainer' programs in industry, she focuses also on techniques which help teach OT and is a main lead in the pedagogical patterns project. She has presented work  in her main areas at ACCU (UK), OOPSLA (USA), OT (UK), XP (Italy) and XP  and Agile Universe (USA).   She is a member of the AgileAlliance ( and a supporter of the Manifesto of Agile Software Development  ( 

Extreme Hour (XH): (workshop) - Jutta Eckstein and Nico Josuttis

Experiencing extreme programming (XP). The XH is "the smallest project in the world", which lasts an extended hour, where people experience most of the aspects of XP, but without actual programming.


Kevlin Henney

Kevlin Henney is an independent software development consultant, which covers the range of sins from development through mentoring and reviewing to training. The focus of his work is in programming languages, OO, CBD, UML, patterns and software architecture.

He is a regular columnist for C/C++ Users Journal, Application Development Advisor and JavaSpektrum, and a former columnist for a couple of magazines that have gone to the wall (C++ Report and Java Report). He contributes to other magazines, including Overload... and other magazines that have gone to the wall (e.g. EXE). He'd like to think that there isn't a causal connection. He is also a regular speaker at conferences in Europe and the US.

It was once said that "a committee is a cul-de-sac down which ideas are lured and then quietly strangled", so Kevlin owns up to membership of the following: committees, boards and panels for the OT conference, the EuroPLoP conference, the VikingPLoP conference, the JAOO conference, Hillside Europe, C++ (BSI and ISO) and C (BSI).

C++ Threading

A lot has been written about multithreading, C++ and multithreading in C++. There have a number of different higher-level threading APIs written and proposed. Some are influenced by object models that are not necessarily appropriate to C++'s own idioms, and some of them suffer from looking too obviously like API wrappers, in spite of their specific API independence.
This talk starts from basic principles to develop a different model for threading in C++. It is simple, idiomatic and generic, and its thinking is more obviously unchained from the view of thread objects as C API wrappers.


Andy Koenig

Andrew Koenig is a member of the Communication Software Research Department at AT&T Shannon Laboratory, which was once part of Bell Laboratories. He has been working mostly on C++ since 1986, and is the C++ standards committee's project editor. He joined Bell Labs in 1977 from Columbia University. Aside from C++, his work has included programming language design and implementation, security, automatic software distribution, online transaction processing, and computer chess. He is the author or coauthor of more than 150 articles and the book `C Traps and Pitfalls,' and coauthor of the books `Ruminations on C++' and `Accelerated C++.' He has taught courses at Columbia and Princeton Universities and Stevens Institute of Technology, given tutorials for Usenix Association, Stanford University, Boston University, Lund Institute of Technology, ACCU, SIGS Conferences, the Federal Open Systems Conference Board, and the Federal Reserve Bank, and given invited talks for IBM, Syracuse University, ACM, IEEE, Miller-Freeman, and AT&T in Tokyo, and a keynote talk for the 10th International Python conference.


The Cost of C & C++ Compatibility [Keynote]

This was first presented as a technical session at the Autumn meeting of WG21 and J16 (C++ Standards Committees). During the talk I will consider various aspects of maintaining backward compatibility within languages whose history extends back almost 30 years. I will be examining the way in which hardware development questions many of the assumptions that are made in language design and use. The main purpose of this keynote is to challenge your view of the C family of languages and ask you to consider how we should move forward.

Parsing made easier - a radical old idea

As powerful as Python regular expressions may be, there are some things they just can't handle, such as nested data structures or sequences of symbols that aren't characters. Snobol4 solved some of these problems 35 years ago, but the language never caught on. In the Python world, the SnoPy project offers a gateway to a Snobol-style pattern matcher that is part of GNAT, the Gnu Ada Translator, but not every platform has GNAT available. This talk describes a pure Python library that captures most of Snobol's pattern-matching capabilities. Moreover, it lets you do some interesting things that Snobol can't do, such as

  • using regular expressions as pattern elements,
  • making it unnecessary to imitate them;
  • specifying how to put strings back together after the patterns have taken them apart; and
  • storing intermediate data in safer places than global variables.

This library is a work in progress, so the exact contents of the talk will depend on how much progress I've made by then. However, I hope to be able to show some examples that are entertaining, and perhaps even useful.


Angelika Langer

Angelika Langer is a freelance trainer/consultant working and teaching in the area of object-oriented and component-based software development  in C++ and Java. She is a recognized speaker at conferences world-wide, co-author of the book "Standard C++ IOStreams and Locales" and author of numerous articles about C++ and Java. A more comprehensive biography can be found at

Concurrent Programming in Java (double session)

Support for programming with multiple threads is a core feature of the Java programming language, yet multithreading issues are ignored by many if not most Java developers. Even innocent classes that do not actively start and manage threads should be prepared for use by multiple threads, which raises issues of thread-safety and synchronization.
In the first ]art of this talk we give a fairly quick refresher of the basics of thread-safety and concurrency control in general and the Java-specific features in particular. We will also address typical problems such as the nested monitor problem and its potential for deadlocks and we will learn how to apply common techniques such as the conflict set method.

Next we tackle some of the more challenging details of the Java thread API. Specifically, we look into thread termination due to an interrupt request or due to an uncaught exception. We will answer questions such as: How does thread interruption work? Why are the methods suspend() and resume() deprecated? What do we do instead? How do we gracefully terminate a thread in response to a thread interruption? What is the effect of uncaught exceptions on active threads? How can we avoid that an uncaught exception prematurely terminates a thread?
Another less commonly known topic is the Java memory model, which has certain problems with atomicity and sequential consistency of access to volatile variables. The Java language specification gives guarantees for volatile variables, but not all Java implementations actually meet these requirements. As Java developers we would like to know which guarantees we can rely on and which areas we should better avoid.

In the second session we discuss patterns for concurrency control and how they can be implemented in Java. We will look into various incarnations of the adapter pattern and will se which kinds of adapters are relevant in concurrent programming (immutability adapters, synchonization adapters). We will learn how to implement a reader-writer pattern in Java, including the before-after-technique. We will talk about patterns for thread completion (like future, callbacks, group proxies).
The tutorial is of interest to Java programmers, but also to developers who implement concurrent programs in related languages such as C++.


Alan Lenton

Alan Lenton is in charge of the design, development and programming of  multi-player games for Interactive Broadcasting Ltd. He also handles  technical matters relating to the delivery of the games over the Internet.

Multi-player games designed include Federation, an adventure/economic  simulation set in a future universe; Iron Wolves, a submarine simulation;  Age of Adventure, a role-playing game based in Victorian times, now in  beta-test; and Barbarossa, a strategy wargame based on the German  invasion of Russia in 1941, now in alpha-test.

He lives in London, and as well as his work for Interactive Broadcasting,  he is a member of the BSI's C++ panel and a long standing member of ACCU.

Alan produces a weekly newsletter covering issues affecting the Internet  and computing generally. Web site:

The Internet, Software and Computers – A Report Card [Keynot]

The demise of the 'dot com' companies, the 'war' on terror, the assault on peer-to-peer technologies, Microsoft's Paladium initiative, the operation of the patent laws and changes to copyright laws have a great deal of potential to change the face of computing as we know it. Desk top computers are a relatively recent innovation, and they are viewed by many powerful grouping in society as a threat to their interests. This talk looks at what is happening as they try to defend their interests and how it will impact those who work with computers.


Simon Peyton Jones

Simon Peyton Jones is a researcher at Microsoft Research, Cambridge, where he has been since 1998.  Prior to that, he spent nine years as a Professor at the University of Glasgow.  His research interests are centered on the design, implementation, and application of programming languages, especially functional languages such as Haskell.  He leads the team that developed and supports the open-source Glasgow Haskell Compiler (GHC).  Home page:

Template metaprogramming in Haskell

In this talk I will draw together two brilliant ideas: the rich but wild world of template meta-programming of C++, and the elegant garden of functional programming.  Functional languages have served as a wonderful laboratory in which to explore and develop language ideas, such as polymorphic type systems, garbage collection, and so on. Template meta-programming has been such a success in C++ that it seems natural to try to transplant it into the context of a higher-order, typed functional programming language, and explore its properties there.

In this talk I will describe how we have done this for the functional programming language Haskell.  I will not assume that you already know Haskell, but I hope to leave you with some idea of what Haskell is and why I love it.   Our Template Haskell project is at a fairly early stage, but the strange perspective of functional programming may perhaps leave you with a different view of templates in C++.


Mark Radford

Studied Maths and Physics at Trent Polytechnic, Nottingham, graduating  in 1986. Having found the computing part of the course more interesting  embarked on a career in software development, using a variety of  languages and platforms - e.g. VAX/VMS, Pick, Windows, Fortran-77, C,  C++ and Java - over the last fifteen years.

Independent since 1997, having initially specialised in Windows  development using C++ and MFC, moved on to specialise more in C++ and  object oriented design working on a variety of developments including  both distributed and embedded systems.

Member of the BSI C++ panel since November 1998, and occasional member  of the UK delegation to ISO meetings.

Pattern Experiences in C++

My first contact with patterns was when I first read the "Gang of Four" book in 1996. Six years later I revisit the GoF book and several other books about patterns on a regular basis. I find patterns a powerful tool in software development, and not just as a source of solutions to problems - seeing applications of known patterns leads to the increased confidence that, in hindsight, problems have been solved in ways already known to be effective!
This talk is about my experiences with patterns - taken from the GoF book and from other sources - over the last six years, and it falls into three parts. First a brief introduction to patterns, and an overview of those participating in the rest of the talk. Second my experiences of learning about patterns. Finally, I will present assorted experiences of implementing and using patterns in my C++ software development work.


Randy Marques

Randy Marques is a CASE-Consultant with AtosOrigin – Technical Automation – In-Product Software. Originally from Curaçao, an island in the Caribbean, he now lives in the Netherlands. His main jobs concern the usage of the C-language, Code Quality Assessments, Coding Standards, Code Architecture and Build Management especially in the area of Embedded Software. In general he tries to fix the gap between SPI (CMM) and the work floor (Programmers) He is also a member of the Dutch and International SC22/WG14 Committee and a teacher of the ‘Safer C’ course developed by Les Hatton.He started with Software Engineering in 1971, he is a real ‘old hand’ with long-term experience of all levels of Software Engineering. The first computer he worked on was an IBM 1410 with 4K of core memory, at that time one of the biggest in Europe.In his spare time he is a father of 3 kids, a diving instructor and on non-rainy days he likes to take his Kawasaki Vulcan 750 out for a spin.

Coding Standards – Given the ANSI C Standard why do I still need a Coding Standard

A few years ago, there was an advertisement of BMW on the British television.
The text was something like: "Contains more computing power than it took to take man to the moon". If you are of the opinion that this is good, you should not miss this presentation.
It is about the flaws in the definition of the C-language, about flaws in Compilers, about flaws in management decisions concerning choice of processors and compilers, about flaws in how programmers handle the language and about what you can do to keep the damage to a minimum.
It will not be a theoretical talk but real-life experience.
Open to anyone with a basic knowledge of C-programming.

Generic Build Support – A Pragmatic Approach to the Software Build Process.

Software Engineering was always so simple: A few directories here, a makefile there. Cup of coffee with the guys.
Now suddenly most of my builds fail, builds are not reproducible and a re-build generates a different executable.  Something wrong with the coffee?
Or are we with too many people trying to build twenty versions of one million lines of code for three platforms the same way we used to build two versions of 4000 lines of code for one platform?
Do we really think that with the experience of building a tool-shed we only need to extrapolate to be able to build a skyscraper?
Do we really need a new dedicated directory- and make-structure for every new project?
Or can we state, in analogy of the construction business, that the build process of a building is always the same irrespectively if you are building a concert-hall, hospital or town hall?
As the title says: a pragmatic approach. No theories. About directory-structures, SCM, makefiles, Compilers and Linkers.
Open to anyone with a basic knowledge of Software Engineering.


Herb Sutter

Herb Sutter ( is the author of more than 160 technical articles and of the widely acclaimed books Exceptional C++ and More Exceptional C++ ( Herb is convener and secretary, respectively, of the ISO and ANSI C++ standards committees, contributing editor and columnist for C/C++ Users Journal (CUJ), C++ community liaison for Microsoft, and former editor-in-chief of C++ Report.

Advanced Template Issues and Solutions (double session)

This in-depth talk assumes you already know the basics about template programming, including specialization, and have a basic understanding of Koenig lookup. It delves into specific nifty techniques you can use and subtle pitfalls you need to avoid, including answers to the following questions:
First, to get our feet wet, we'll look at a not-well-known corner of Standard C++ templates: What are dependent names, why does the standard mandate two-phase template name lookup in the first place, and how does the two-phase lookup work? Most importantly, how can you make it work for you, and how can you account for the fact that today's compilers do the two-phase lookup variously well?
Second, a look at a cutting-edge issue that has only recently been discovered and debated among the inner circle of world's top C++ library writers, about writing templates safely: Why do you sometimes want and sometimes not want to have Koenig lookup work in your templates, and how can you turn it off when you don't want it? The issues involved can be subtle, they affect your templates today, and they have only this year become better understood.


Alan Griffiths

Alan Griffiths ( is a long-standing contributor to the ACCU journals and mailing lists and chair of the ACCU. He is also a member of the BSI C++ Panel (although not a very active one). In working life Alan ( is "Software Development Champion" at Microlise Limited. He has responsibility for introducing both developers and management to more effective ways of developing computer software. For most of Alan's technical articles and various other goodies visit:

Agile Enough?

The group of development methodologies recognised under the banner "Agile" have gained a lot of attention recently. They are clearly successful for a range of problems, but is this range broad enough?
My experience is that there are many projects for which "key" process elements cannot be followed, for technical, business or organisational reasons. Indeed, most of the projects I've worked on fall into this category.
How important are these types of projects? Can they really be forced into an "Agile" straitjacket? Can they be delivered effectively?
Is "Agile" sufficiently liberal in scope?



Sarah Lees

In working life Sarah ( is "Software Architect" at Microlise Limited. Having been introduced to more effective ways of developing software by Alan Griffiths, she has been working with Alan in improving the way that software is developed at Microlise. Having been with Microlise since 1994 she is familiar with the way in which the company has grown and changed over the last 8 years.

The Organisation Strikes Back – Alan Griffiths & Sarah Lees

In "Reworking the Organisation" (presented at the ACCU Conference in 2002) Alan described his first six months at a new employer in the form of a case study.
While there were some early gains there were a number of import questions unanswered: What would happen next? Were the changes dependent on Alan's continuing involvement? How would management intervene? How well would the changes be internalised by the organisation?
In "The Organisation Strikes Back" Alan and Sarah extend the case study to cover a two-year period and (plus some of the history) and, in the process, answer these and other questions.


Walter Banks

Walter Banks is the president of Byte Craft Limited, a company specializing in software development tools for embedded microprocessors. His interests include highly reliable system design, code generation technology, and programming language development and standards. Walter Banks is a member of the Canadian delegation to ISO WG-14, where he co-authored WDTR 18037 (a technical report on C language extensions to support embedded processors). He has co-authored one book, and numerous journal and conference papers.

Linguistic Variables: Clear Thinking with Fuzzy Logic

Linguistic variables represent crisp information in a form and precision appropriate for the problem. For example, to answer the question "What is it like outside?" one might observe "It is warm outside." Experience has shown that if it is “warm” and the time is mid-day, a jacket is unnecessary, but if it is warm and early evening, it would be wise to take a jacket along (the day will change from warm to cool). The linguistic variables like “warm”, so common in everyday speech, convey information about our environment or an object under observation. We will show how linguistic variables can be defined and used in a variety of common applications, including home environment, product pricing, and process control. The use of linguistic variables in many applications reduces the overall computation complexity of the application. Linguistic variables have been shown to be particularly useful in complex non-linear applications. Linguistic variables are central to fuzzy logic manipulations, but are often ignored in the debates on the merits of fuzzy logic.


JC van Winkel

JC van Winkel has a B.S. and an M.S. in computer science (the M.S.from the Vrije Universiteit Amsterdam). He works at AT Computing, a small courseware and consulting firm in Nijmegen, the Netherlands. There he teaches UNIX and UNIX-related subjects, including C++.  He has presented C++ tutorials at nine OOPSLA's since 1993.

GNIRTS ESAC REWOL - Bringing the UNIX filters to the C++ iostream library.

The UNIX technique of chaining filters to manipulate streams of data can now also be used in C++ in combination with the iostream library. A framework allows users to build ostream objects without having knowledge of the iostream structure. These ostream objects build upon existing ostream objects, but manipulate the data being output. 
Examples of these filtering techniques are: an uppercase/lowercase filter, an ``logging'' filter, that precedes every line with a timestamp, and a ``tee'' filter that allows multiplexing output to several ostreams.


Ulrich Eisenecker

Dr. Ulrich W. Eisenecker is a professor for computer science at the University of Applied Sciences Kaiserslautern, Zweibruecken. He spent nearly a decade in industry and closely cooperates with industry today. For some time he was national agent of ACCU in Germany. He edits KOMPONENTEN-Forum, a regular supplement of the journal OBJEKTspektrum, dedicated to component-technology, and co-authored with Krzysztof Czarnecki the book "Generative Programming: Methods, Tools, and Applications" published by Addison-Wesley.

The Roadmap to Generative Programming With C++

Today most applications are developed as individual systems. Very often variants of these systems are required eventually, because of the needs of different customers and diverse contexts of use. Thus, following an approach for engineering software system families from beginning is of significant advantage. Generative Programming is a software engineering paradigm based on modeling software system families such that, given a particular requirements specification, a highly customized and optimized intermediate or end-product can be automatically manufactured on demand from elementary, reusable implementation components by means of configuration knowledge.
Because of its specific implementation of generic polymorphism, C++ templates offer a unique way for static metaprogramming. This allows the implementation of advanced compile-time configuration generators that are at the heart of Generative Programming.
After introducing the fundamentals of Generative Programming, techniques for implementing domain specific languages, configuration generators, and elementary implementation components in C++ will be highlighted.
Finally, I will present an overview of how to add powerful genericity to non-template languages using a frame-processor thus enabling Generative Programming.


Hubert Matthews

Hubert is a freelance software consultant and trainer based in Oxford. His main areas of expertise are architecture and design and he has been an architect for a number of clients, both small and large.  He has also given training courses across the globe in a range of subjects including C++, Java, UML, EJB, OOA/D, patterns and components.  Other areas of consultancy include technical strategy, architectural audits, and process issues.

Following a degree and doctorate in engineering at Oxford (LMH) and a student apprenticeship with GEC in Rugby, Hubert worked at CERN in Geneva before returning to Britain to indulge his other passion: singing.  Since leaving the Royal Northern College of Music in Manchester, Hubert has continued to juggle his two careers as a professional concert and opera singer and as a freelance software consultant (including the obligatory dot-com nightmares and other abortive get-rich-quick schemes).

In his abundant spare time, Hubert coaches rowing, dances salsa and drives too fast.

The Timing and Cost of Choices

As programmers and designers, we are contantly make choices. We think very hard about these choices but how often do we think about the choice process itself? What are the consequences of choosing now versus delaying, and when do we validate those choices? This talk compares the effects of early versus late binding for code, process and requirements and shows that there are patterns that relate: product choices and process choices, binding times and the timespan of choices, the effects of being able to revisit those choices later, and the choice of emphasis on error prevention, removal or tolerance.

The consequences of binding time choices Choose, Check and Use pattern Product v. process - feedback and metrics Cost of failure and error prevention v. appraisal v. tolerance


Steve Clamage

Steve Clamage has been involved in C++ compilers since 1980. He is responsible for C++ compiler development at Sun Microsystems. He is chair of J16, the ANSI C++ Commitee, and moderates the comp.std.c++ newsgroup.

The Stability of the C++ ABI

As C++ evolved over the years, the Application Binary Interface used by a compiler often needed to change in order to support new or changing language features. Programmers expected to recompile all their binaries with every compiler release. An unstable ABI is incompatible with the Solaris philosophy of shared libraries, and is a nightmare for library and middleware vendors. With the advent of the C++ Standard in 1998, there is new hope for a stable C++ ABI on Solaris platforms. This paper addresses the main issues for ABIs using our experiences with Sun C++ on a Solaris Platform.


Chris Hills

Embedded Engineer. My background covers most types of non-PC based programming from Unix systems to smartcards and various forms of electronics. Currently the BSI convenor for IST/5/-/14 (the BSI's C Standard's Panel to you and me!) Also a member of the MISRA-C working group. Past papers presented to the ACCU are at http://

What can MISRA-C (2nd Edition) do for us?

MISRA-C a small coding guide for the UK Automotive industry has escaped not only from the automotive industry but is now used worldwide. It has sold at a rate of over 1000 a year in the last 4 years!
This paper will look at why MISRA-C was needed, what it gives us and what is new for the second edition. The new version of MISRA-C will be published in March 2003 but as I am on the working group I can give an insight into the changes and the process used to produce what will be the new version.


Gabriel Dos Reis

Former student of the École Normale Supérieure de Cachan (France), Gabriel Dos Reis got his PhD in Mathematics (Differential Geometry) at Université de Paris VII in 2001. His research interests include applying numerical and computational methods in Geometry and especially in the construction of constant mean curvature surfaces. He began lobbying for C++ in scientific computations since 1996 when he worked on the European scientific project FRISCO (Framework for Numerical and Symbolic Computations) at INRIA Sophia Antipolis (France). He is the author of the numerical components of the new GNU libstdc++-v3. Currently, he is holding a post-doctoral position at the GALAAD project of INRIA Sophia Antipolis. He is also the co-maintainer of libstdc++-v3.

What is the Type of std::toupper()?

The template-argument deduction facility found in standard C++ template machinery heavily depends on the ability ro ascribe types to expressions. That some expressions in C++ don't have (native) types can unncessarily complicate ways in which some natural expressions may be written in a generic framework. This talk explores some points in the space of paths connecting functions and function objects, all in the framework of polymorphism.


Barb Byro

Barb Byro is a project manager who has spent the last 3 years managing Telecom network and collocation projects for customers of Metromedia Fiber Networks in both New York and London.  She's also lead ISO compliance teams, devising practical strategies to make ISO into a useful tool rather than a time wasting, irrelevant exercise. Barb's also managed online computer game software development projects for over 10 years. In that capacity she's worked with paid and volunteer staff, both in-house and external.

How to Handle Project Managers: a survival guide

A practical guide to project managers. Having a good working relationship with your project manager can make or break a project. We'll discuss strategies to evaluate and work with project managers. Not all project managers appreciate programmers, but they can be trained.  We'll also touch on how this can make ISO compliance less onerous.


Lois Goldthwaite

Lois Goldthwaite  has been programming professionally since 1983, specialising in C++, C, and Java. As convenor of the BSI C++ panel since 1997 she represents the UK at international standards meetings. She is also convenor of BSI's Posix and C# panels, a member of the C panel, and ACCU's Standards Officer. After a lengthy period of steady employment, she has returned to being a computer mercenary. For recreation she takes out her aggressions on the squash court.

Fun and Functionality with Functors

Functors, or function objects, bring power and flexibility to C++'s Standard Library, but their contribution is too often overlooked or underestimated. This talk explains what function objects are, how to create and use them, and what pitfalls to avoid. In passing, there is also some discussion of ways to achieve similar functionality in other programming languages.


Arno Schmidmeier

Arno Schmidmeier is an independent consultant focusing on aspect oriented software development and enterprise application integration. During his very successful time as Chief Scientist at Sirius Software GmbH, he was responsible for the commercial adoption of new technologies like AOP.   He deployed several large-scale projects based on AspectJ.   He offers consulting services for use and introduction of AspectJ in commercial projects.
He was an independent expert on JSR 90.

Aspect Oriented Programming for Enterprise Application Integration

Aspect oriented programming is a new but urgently needed software development paradigm, to modularize tangling code at all levels.
Aspect oriented programming is widely used for low level elements like synchronizing, tracing, logging, exception handling, etc.
But it also unleashes its power in EAI-projects, as it has proofed in several very successful projects.


Julian Smith

I live in Oxford in the UK, where I work as a contractor. I've worked for various companies in the last 6-7 years, did a PhD in Cognitive Psychology at Edinburgh University, and studied Physics as an undergraduate at Oxford University. I've written Cmm, a C++ language extension/translator that adds Multimethods to C++, and I wrote an article about Cmm and Multimethods in Overload, April 2001. Other info is at

C++ & Multimethods

I think that language support for multimethods could simplify some common programming tasks, and that the lack of support for multimethods in most programming languages has lead to inferior ad-hoc solutions being adopted and standardised, with little recognition that there can be a better way. A couple of examples are GUI event dispatching (Win32 message maps, Qt signals/slots), and internationalisation support. Multimethods also enable freedom from rigid interface definitions (for example they allow the equivalent of virtual functions to be added to any existing class); this freedom can be misused, but is sometimes very useful - consider how useful traits classes are when used to embellish existing types. Finally, I may talk about how multimethods can be used as part of a TeX-style typesetting engine, if some experiments in this area workout.


Nathan Sidwell

Short bio:Nathan is a consultant at CodeSourcery, and has been involved with G++ for around 4 years. Previously he has worked as a processor architect, before falling into compilers, and then teaching at the University of Bristol. Prior to that he pretended to be a physicist, and knows how to use a soldering iron.

A common vendor ABI for C++ – GCC's why, what and what not - Nathan Sidwell and Gabriel Dos Reis

The itanium C++ ABI has been adopted by a number of compiler vendors to provide compatibility between itanium compilers. The generic parts of the ABI can be used on other architectures, and GCC took the opportunity to revise its C++ ABI for all targets. The previous ABI had grown by agglomeration as C++ matured. Now there is a complete standard a better ABI can be designed from the ground up. However, common practice can restrict what can be done. The talk will cover both how the ABI came about, some of the design decisions, how it can affect user programs, the promises it makes, and the promises it does not make. We'll also demonstrate concrete use of the ABI to build libraries or tools for cross-modules, cross-compilers, or cross-language systems.


Brandon Bray

Brandon Bray is the program manager for the Microsoft Visual C++ compiler front-end and language team. Prior to working at Microsoft, Brandon came from Cornell University where he concentrated in advanced programming language design and compiling for high performance architectures.

Inside Security Checks and Safe Exceptions

Buffer overrun attacks continue to be the leading security issue facing the software industry. In this talk, you will learn the details of specific methods attackers use to exploit buffer overrun vulnerabilities, from inserting arbitrary code to hijacking function pointers to hijacking the exception handling mechanism itself. Using Visual C++ as a concrete example, you will learn how C++ compilers can reduce or even eliminate different classes of security attacks, how the VC++ implement security checks and safe exceptions, and how the mechanisms work to make software more robust against attack.

Design Patterns in C++ and C# for the Common Language Runtime

Writing code for the common language runtime necessitates a certain number of considerations. Among them are efficiency in the face of a managed execution environment, verifiability, and library portability between languages. This talk will address some of the frequent design patterns used by programmers familiar with C++ and C#.



Studying at a Distance

The objective of this panel is to explore, with the audience, ways in which 'distance learning' can be made to work effectively. The main thrust of this panel will be considering what students can do to improve the value they get from learning this way.
The growth of the Internet, email and other forms of electronic communication has added new dimensions to the traditional mechanisms of distance learning. However the drop out rate continues to raise questions as to how students can get full benefit from such study methods.


Being a Mentor

This panel will focus on all aspects of mentoring, both in person and remotely as part of a distance-learning project. The panel will draw on a wealth of personal experience to challenge the audience to consider how they can become mentors, and if they already are, how they can better meet the needs of those they are guiding.


Goldfish Bowl

The Future of Programming Languages

Changes in hardware have challenged language designers to consider adding new features to languages. On the other hand, issues of backward compatibility make changes to languages unpopular.
In this interactive session the audience under the guidance of a 'management team' will discuss ways of resolving the tension between the needs of the future and those of the past.
Guidance will be given at the start of the session to those who have not previously experienced this type conference item. You will be able to remain in a purely observational role if you wish but opportunities for active participation are an essential ingredient of a 'goldfish bowl.'


Jeremy Siek

Jeremy Siek is a Ph.D. student at Indiana University Bloomington, studying programming languages in the Open Systems Laboratory directed by Dr. Andrew Lumsdaine. Jeremy is a coauthor of several libraries that have become part of Boost, including the Boost Graph Library and Boost Iterator Adaptor Library. Jeremy is a coauthor of the book "The Boost Graph Library: User Guide and Reference Manual". His master's thesis, the Matrix Template Library, applied generic programming to numerical linear algebra. On the industry side of things, Jeremy did a one-year internship under Alexander Stepanov in the SGI C++ compiler and libraries group, and a summer internship with Bjarne Stroustrup at AT&T Research Labs.

Design and Implementation of the Boost Graph Library

The Boost Graph Library (BGL) offers an unprecedented level of flexibility and power to its users due to its generic programming and generative design. This talk describes the motivation for the core abstractions of the BGL. We will take a close look at a key family of graph algorithms, and how generalized version of these algorithms came to be written as function templates in the BGL. We then will switch focus to the graph data structures provided in the BGL, and look under the hood of the adjacency_list class to see how generative programming techniques were applied to create this Swiss-army knife of a graph class.


Guido van Rossum

Guido van Rossum created Python, a major open source programming language, in the early 1990s at CWI in Amsterdam. He is still actively involved in Python's development. In 1995 he moved to the United States, where he now lives and works in Reston, Virginia for Zope Corporation -- developers of Zope, the leading open source content management system (written in Python, of course).

Why design another programming language? [Keynote]

The creator of Python muses on the forces that influenced Python's design, and what lies ahead.


Duncan Booth

Duncan graduated as a Computer Scientist from Cambridge University and initially worked for Torch Computers using BCPL, C and Assembler before joining the fledging RCP as the third employee in 1984. Duncan has worked on many projects in RCP, including foreign exchange dealing, real-time financial data, banking back office systems, and healthcare systems. His interest in Python was first sparked at a science fiction convention, since when it has become a major player in his choice of languages. As well as writing in Python he also ported it to the Psion 5 handheld. He is currently working with a variety of internet based systems including Python and Zope.

Python Design Patterns

What design patterns are applicable to Python? Some patterns are an intrinsic part of Python, other patterns require some careful coding to get the best from them. What new patterns appear in Python? This slot will also include time for anyone wishing to discuss, explain or demonstrate their own patterns.

Integrating Python, C and C++

Sometimes Python on its own isn't quite enough. Maybe you want to access a library available only in C or C++, maybe you want to script an existing C/C++ application, or maybe you just want an easier way to test your C/C++ code.

This talk introduces the Python C API, and also looks at more sophisticated interfaces such as SWIG, SIP and Boost.


Duncan Grisby

Duncan is an experienced software architect and developer specialising in distributed and concurrent systems. He is the lead developer of omniORB, a high performance open source CORBA implementation. From 1999 until its closure in April 2002, Duncan was a researcher at AT&T Laboratories Cambridge, Europe's leading computing and communications research laboratory. Since leaving AT&T he has been working as an independent contractor on a number of large-scale distributed systems integration projects. Duncan holds MA and PhD degrees from Cambridge University, England.

CORBA? Isn't that obsolete?

If you keep up to date with the mainstream IT media, you could be forgiven for thinking that CORBA was obsolete, and thoroughly dead. In fact, it is alive and well, and is the best solution to a wide range of distributed application problems. This talk introduces the features of the CORBA object model and services, and discusses when it is most sensible to use CORBA, in comparison to other technologies. The talk is based around using CORBA from Python, but it also shows how it can be used to build robust distributed systems with a mix of programming languages, including Python, C++ and Java.


Andy Todd

Andy has been involved in the PythonCard project since its inception in 2001. As one of the developers on the project he has contributed code, samples and documentation. His areas of special interest include storage, in particular relational databases, and object-relational mapping as well as agile methodologies.

Building GUI Applications with PythonCard and PyCrust

In this session, we will demonstrate the process of building full-blown Python applications with world-class graphical user interfaces. By giving you access to the full power of Python without requiring you to master complex GUI libraries, PythonCard becomes a power tool for scripting users and professional programmers alike. Creating graphical desktop applications that run across platforms and take advantage of Python's simple elegance is easier than you think. In addition PyCrust, the embeddable Python shell, puts the power of the Python interpreter at your fingertips. That means every GUI widget in your application can be inspected and manipulated from the PyCrust shell while your application is running. The combination of PyCrust and PythonCard brings the power and flexibility of Python to the realm of cross-platform GUI applications.


Marc-André Lemburg

Siena Web Service Architecture

In today's business world, Web Services authoring has become a synonym for having to write JavaBeans-based Java code for deployment on J2EE compatible application servers. This talk will demonstrate an alternative solution which is built around Python with the aim of making service writing easy, flexible and productive.


Chris Withers

Chris Withers has been developing Python and Zope applications for New Information Paradigms ( for the last three years. He is the maintainer of the Squishdot project as well as doing core development work on both Zope and its Content Management Framework. His areas of special interest include unit testing and content management systems. 

Extreme Programming in Python

This talk will introduce the basic principles of Extreme Programming, and explain Python's support for unit testing, including "unittest" and associated tools. It will draw on NIP's real world experiences of using this methodology in production. In addition time will be provided for anyone else working with XP to recount and discuss their experiences and demonstrate associated testing tools.


Hamish Lawson

Hamish Lawson is a software developer in the IT Services department at the University of St Andrews, Scotland. He specializes in web-based enterprise information systems.

RESTful Python

Web Services are all the buzz just now - and for many people, Web Services means SOAP and other new protocols. But a number of notables from the Web and XML worlds are concerned that, in adopting an essentially RPC model rather than a resource-based one, SOAP discounts the architectural lessons that have been learned about what made the Web successful - so making it harder for SOAP-based Web Services to achieve the ubiquity enjoyed by the Web itself. Instead it is possible to deploy Web Services that do follow the Web's architectural principles and that use the established Web protocols we already know. These architectural principles have been captured under the name Representational State Transfer, or REST. In the first half of my talk I will discuss what these REST principles are and how they help in creating maintainable and interoperable web spaces. In the second half I will survey the support for REST provided by various Python web application frameworks, in particular Quixote.


Alex Martelli

Alex Martelli lives in Italy and is a Senior System Developer with AB Strakt in Sweden, mostly writing Business Logic infrastructure and modules for Strakt's CAPS Real-Time Enterprise Computing platform. Alex co-edited the "Python Cookbook", and wrote "Python in a Nutshell", both for O'Reilly. Alex is a popular poster to comp.lang.python, the winner of the "Python Activators' Choice 2002" award, a member of the Python Software Foundation, and a board member of the Python Business Forum.

Introduction to Python and Jython for C++ and Java Programmers

A general 90-minutes tutorial about Python and Jython for experienced C++ and Java programmers attending the rest of the ACCU conference


Anthony Eden

Anthony Eden is the President and Lead Developer at Aetrion LLC which was formed to provide development support for organizations utilizing open source projects. Anthony has been a Java developer for over 7 years and a recent Python crossover. Anthony develops numerous open source projects, including JPublish, a web publishing and application framework and the Open Business Engine, an open source workflow engine.

Scripting Java Applications with Jython

Java is currently one of the most powerful server-side languages available. Thanks to the fact that it is cross-platform, dynamically linked and has a very comprehensive set of standard and open source libraries, Java is a very good solution for server-side development. The main problem with developing server-side Java applications is the edit/compile/deploy portion of the development cycle which increases development time. Enter Jython. Since it is written in Java, Jython acquires all of the benefits of the Java platform without the need for a compile cycle. In this presentation I will explain how the JPublish web framework ( ) relies on Jython as the language to glue together the view and the model of an MVC framework to reduce development time dramatically.


Paul Brian

Lightning Talks

This session will consist of a number of 5 or 10 minute talks, as held to wide acclaim at recent Python events.
Paul will accept entries during the conference. And since many other topics are too short for a 90 minute talk, we may well fit more lightning talks into other slots too!


Jacob Hallén    

The Infinite Filing Cabinet - object storage in Python

Editor's note - within a 90 minute slot we intend to provide a general survey of Object Storage techniques in Python. This will hopefully include a full talk on ZODB (speaker volunteers welcome) for use outside of Zope, as well as the following talk which will last 30-45 minutes: The workhorse of the CAPS system is a database and supporting software which is called the Infinite Filing Cabinet, the IFC. Information is organised as objects and the IFC, written in Python, is the persistent storage of these objects.Objects can be searched for, retrieved and dynamic notification of object creation or modification can be requested. More complex actions can also be performed when objects transit specified states at optionally specified times. The IFC does not support the deletion of data; this is performed by deprecation. It is therefore possible to access any object in any of its prior forms.


Jaako Jarvi

Jaakko Järvi has a Ph.D. in computer science from the University of Turku, Finland. He is currently a post doctoral researcher in the Open Systems Laboratory, one of the Pervasive Technology Laboratories at Indiana University. His research interests include programming languages and generic programming. Besides academia, Jaakko has worked as a software engineer at ABB Corporate Research, and as the CTO of Atuline Ltd., a university spin-off company on its way to maturity. Jaakko is the principal author of the Boost Tuple and Lambda Libraries and participates actively to the C++ standardization work.

The Lambda Library : Unnamed Functions for C++

More often than not, small and simple function objects are needed solely to be passed to an STL algorithm, having no further use in the program. Defining new functions or classes for just this purpose is verbose and adds unnecessary names to the program. The STL programming style would greatly benefit from lambda abstractions: unnamed functions that can be defined where they are used, at the call sites of STL algorithms. As unnamed functions are not part of core C++, the standard library provides
binders, adaptors and the set of elementary function object templates (less, negate, ...) as an alternative. However, these tools have many restrictions and leave much room for improvements.

The Lambda Library (LL) offers these improvements, providing a rich set of tools for defining unnamed functions, with an intuitive syntax.

This talk explains why LL makes using STL easier, simpler and more fun.
The talk describes the main features of the LL, and outlines its design and implementation.


Walter E Brown

A C++ programmer for over half of his 40 years in computing, Walter Brown earned a B.A. in mathematics and M.S. and Ph.D. degrees in computer science.  His professional career has encompassed academic, industrial, and research environments.  He is currently in the Advanced Physics Software group at Fermi National Accelerator Laboratory near Chicago, Illinois, where he functions as an in-house C++ consultant and educator.

Dr. Brown is also an active member of the ANSI C++ Standards Committee.  He most recently authored "A Proposal to Add Mathematical Special Functions to the C++ Standard Library" in an effort to improve the usefulness of C++ in the scientific and mathematical user communities.


C++ Metaprogramming Concepts and Results 

(Note: suitable for those new to metaprogramming and templates)

Programming with templates in C++ permits program design and coding at significantly higher levels than previously possible.  Such "metaprogramming" can lead to efficient, configurable code that can be reused in a number of contexts.

Starting with basic concepts of parameterization, we introduce C++ templates.  As each language feature is presented, we apply it in the context of small, useful meta-algorithms.  Along the way, we describe the role of the compiler in processing and adapting such code, and provide some representative measurements.

We also present numerous examples of template-based programming practices and techniques that have been developed in recent years.  Topics include generic programming, traits, concepts, adaptors, type computations, and compile-time control structures.


Marc Paterno

Marc Paterno is a physicist-turned-software designer working at the Fermi National Accelerator Laboratory (Fermilab), near Chicago, Illinois. He earned a Sc.B. in physics from MIT and and M.S. and Ph.D. degrees in physics from the State University of New York at Stony Brook. He worked for ten years on high-energy physics experiments, with much of his time spent developing customized software for the analysis of multi-terabyte data samples. 


He has been programming in C++ since 1991, and has been involved since 1994 with the production of object-oriented systems at Fermilab. He has contributed to the design, development, implementation, and review of C++ class libraries used by an international collaboration of more than a thousand physicists in several of Fermilab's largest experiments, in the collection and analysis of more than 100 terabytes of data per year. He is a participant in the ANSI C++ Standards Committee.


Defective C++

Many fields of human endeavour share a common approach to the learning of proper technique: the study of examples.  It is traditional to examine instances of both good style to be emulated as well as poor practice to be avoided. The skilful critique of imperfect art – critical analysis – is a powerful technique to improve the quality of one's own work. 

This talk presents case studies of poor practice in C++ programming. Each example is excerpted (with only some light editing) from actual code developed by professionals. Each is followed by detailed discussion of the code's pitfalls, and concludes with a list of lessons to be learned.