Together: supercharging software development, from start to finish!
Together delivers a superbly integrated set of lifecycle tools for building software with breakthrough quality, efficiency, and performance. Now there's a definitive guide to making the most of Together: Better Software Faster. In this book, two leading Together mentors share the techniques you need to succeed every step of the wayfrom planning and requirements through development, debugging, delivery, and administration. You'll find detailed examples and practical insights for:
Whether you're an analyst, architect, designer, developer, or manager ... no matter what enterprise technologies you're already using ... this book will help you streamline and supercharge your very next software project ... Together!The Coad Series: Improving the ways people work together
Every book in the Coad Series (Peter Coad, series editor) delivers practical guidance for building better businesses, and the systems that support themfaster, better, and more easily. Legendary software methodology expert and TogetherSoft founder Peter Coad personally selects authors and books for this series, and works on a strategic level with every author throughout the book development process.
(NOTE: Most chapters conclude with What's Next.)
1. Together—The Difference It Makes.
We Need It Now! The Principles Behind This Book. Why Is Together Exciting Technology? Process, Process, All the Way.
The CarServ System. Evolving the System. Inspecting the Single Model. Change and Redeploy. Documentation Generation.
Elements of Specification. Domain Modeling. Types and Classes. Modeling Types as Classes.
Business Process. System Functionality. Modeling the Users: Actors. Clarifying Requirements with State Chart Diagrams. Nonfunctional Requirements. Configuration Management.
Getting in Control. Why Feature-Centric? Why Timeboxes? Why Adaptive? Estimating the Costs of Features. Use Cases Versus Features.
How to Measure Quality. Testing. Metrics. Audits. Generating Documentation.
A Worked Example. Effective Interaction Design. Effective Class Diagrams. Using Patterns. Refactoring With Together. Customizing Together.
What Is Architecture and Why Is It Important? Framework First, or Functionality First? Responsibility for Architecture. Documenting Architectural Constraints. Managing Dependencies. Interactions Between Tiers. Managing Versions and Configurations.
Using Together Makes it Easy. It's Not That Easy.
We Need You to Do it Again! A Simple Summary. Now Over to You.
Before You Start. Together Download and Install. The Case Study. Quick Test.
Motivation. Getting Started. Using Together's Testing Framework to Create TestCases and TestSuites. Behind the Scenes. Extensions. Writing Tests in Practice.
Motivation. Tips and Tricks. Bean Properties. Documenting Pattern Instances. Colored Notes. Before-and-After Object Diagrams. Documenting Package Dependencies.
Motivation. Behind the Scenes. Collections API Templates. Templates.
Motivation. Inspector Property Builder. Using the Config-Based Inspector. Open API. Inspector Framework. To Conclude.
Motivation. Framework Classes. Possible Enhancements.
There is a vast difference between a product that forces you to change the way you work and one that inspires you to work differently. Being forced to do it "their" way is uncomfortable and usually unproductive. The freedom to discover better ways of working is more enjoyable and inspires new insights into other approaches to get your work done.
The reason we wrote this book is a product called Together ControlCenter and the different way of working that it enables. We didn't just write it for users of Togetherusers of competing development platforms or no development platform will, we hope, also find it helpful, but we admit, Together is the inspiration behind the book.
You won't often come across products that fit into the "inspirational" category, but it seems for many users that's exactly where Together fits. It is certainly revolutionary compared to the typical modeling and code-generation tools that preceded it. For us, Together ControlCenter has caused us to look again at how teams work together. So, one of the objectives of this book is to show you how Together will help you to develop software in ways that deliver the same or better quality software in a shorter timeframein short, how to make better software faster.
We wrote this book for software development teams, their team leaders, and their managersparticularly, small teams using Java or similar object-oriented languages as their programming language.
Small teams consist of more than a pair of developers and fewer than a dozen. A small team doesn't need subdivisions for management purposes; it has minimum administrative overhead and it consists of people who generally all know each other wellat least they do after a few months of the project! That gives small teams great advantages over large teams, even though their resources are less. The likelihood of a project failing increases significantly with the team size. This is a good reason why a great deal of commercial software (if not the majority) is developed by small teams.
Small teams don't necessarily need to use a named software development process. Other books in this series are designed to help the practitioners of various named processes, such as FDD (Palmer & Felsing 2002), XP (Astels, Miller, & Novak 2001) and UP (Norman & Kranz, in press), to successfully apply them. This book, however, focuses more on recommending the simplest way to get things done while drawing on many of these more formal processes for their best ideas and advice. Our aim is to give development teams maximum help with minimum overhead.
This book is written for those using Java to implement their software. We hope this won't totally deter those using C++, C# and VB.NET especially, from seeking insights in its pages, but we must apologize to them now that throughout we've used Java examples, our downloadable case study is in Java, and discussion of distributed systems and persistence draws exclusively on Java environments, especially the J2EE architecture.
We hope, of course, that this book will also be of interest to many others: those in larger teams or using other languages; those involved as business analysts, users, or testers; and indeed to students, teachers, trainers, and consultants. However, the book is for a technical (software development) audience, not a non-technical one. The best projects tend to be those in which there is a trust between the business and the technical staff as to what can be accomplished in any given timeframe. In our experience, this trust is best built up when the business representatives and users take real interest in the development process, even in some cases joining as full members of the team. We do discuss areas where business and user representatives actively participate in the development process, particularly modeling the problem domain and specification and prioritization of the requirements. In these areas and others, we hope the book contributes to increased understanding between the technical and non-technical teams. So, if you think of yourself as primarily non-technical, then we hope you learn something about what Together can do and how software can be developed more effectively by using it.
In Chapter 1, "TogetherThe Difference It Makes," we talk about why Together is an exciting technology. We are not interested in the marketing flannel here. Together does something different from other development platforms, and that opens up new possibilities. We want to look into what this is and why it can make a significant difference to a team's performance. Also in Chapter 1 we introduce four foundational themes that run through most of the other chapters:
Maintaining just one single-source model is based on the idea that information about the system should be stored in only one place and maintained only once, even though it is viewed and modified in many different forms.
The minimum meta-model is a view of the essential elements in any development process that need to be defined and effectively cross-referencedrequirements, tests, design and implementation.
The perturbation change model is a way of looking at the development process in terms of moving from one valid build to the next in small iterative steps. This is not a different development process but a way of looking at most iterative evolutionary development processes.
Continuous measurement of quality is a necessary requirement for iterative evolutionary lifecycles, that put different stresses on the development activities and the tools environment compared to waterfall-style lifecycles.
After this introduction, what about the structure for the rest of the book? Many books are structured by each step in a development process from requirements to deployment, or by taking each UML diagram type in turn. But there are some snags with that approach. First, it is hard to relate the diagram types to a software development processmany diagrams are touched at different points in the software development process and with a different emphasis at each point. Second, it is artificial. Together invites you to develop your software in an organic and evolutionary way, and we wanted to write a book that invites you to read and explore similarly. Another pressing reason we decided against structuring the book by the different diagram types is because it would be rather dullfor you to read and for us to write!
Then, we were posed an interesting question: What's the least that needs to be done to deliver and use software? Well, if you have some software already (and there nearly always is some), the answer is to just deliver and use that software. You may do other steps before that, particularly if you want different or additional functionality or quality, but you must also deliver the software. Furthermore, in looking for a common way of thinking about the development process for small teams, this last step is also the starting point, since before you can add functionality to a system, you must be sure its current functionality works.
So, we decided that Chapter 2 should start herethe "last" step. That chapter is called "The Last Step: Deploy and Run!" We take the opportunity to introduce our example domain model that will be used for most of our illustrations. We also explore many of Together's capabilities with a system that actually runsa good way to learn. The best way to read this chapter is alongside your computer where you can access the software discussed at this book's companion web site. If you have access to the Web now, check it out at www.bettersoftwarefaster.com. By starting here, we can also ensure that those of you using the book with this software are starting from a valid starting pointone where the existing code and the chosen development environment run correctly and the defined tests all pass.
In Chapter 3, "The First Step: Model the Domain," we look at an important launching point for a new team. While this is also a good opportunity to discuss the essentials of the class diagram, the main emphasis is on how to build a model that will act as an effective interface between business experts who know the essence of the business and the vision for the new system and the software development team that must realize the vision. This model is also important, as it defines the vocabulary for both requirements and implementation classes, and it links too to the definitions of persistent data and user interaction. This chapter uses object modeling in color, as defined by Peter Coad and others, which helps to bring the analysis patterns and domain-neutral models to life and enhance the readability of the models (Coad, Lefebvre, & De Luca 1999).
Gaining a good understanding of the requirements is sometimes overlooked by small (and larger) teams in time-pressured projects, but this is a potentially project-fatal mistake. In Chapter 4, "The Stakeholder Step: Specify Requirements," we discuss capturing specifications using UML, particularly the use case and activity diagrams. The diagrams are simple to understand, which is essential to make them effective at this level. But building effective requirements models is by no means easy, and we discuss key issues to keep in mind, especially strategies for keeping the requirements simple and amenable to management in an agile, iterative process that small teams need.
Chapter 5, "The Controlling Step: Feature-Centric Management," is an explanation of a planning, estimating, and project control process that we recommend to teams. Our interest here is to keep planning and management as simple and as transparent as possible. Again, the question posed before is relevant: What is the least we need to do to deliver the software?that's our goal. One key simplification introduced to improve the management of iterative lifecycles is to merge the units of planning with the units of requirements, whether these are use cases or features. We refer to development processes that make this simplification as "feature-centric"hence the title of this chapter.
Some might interpret the goal of seeking the least that is needed as suggesting that we are looking for the line of least resistance, and inevitably the quality of the software will suffer. While this could be a danger, it is certainly not our meaning nor our intention. Chapter 6, "The Continuous Step: Measure the Quality," looks at how to ensure that quality is maintained throughout the development process. Our book is called Better Software Faster, to which the natural response is "Better than what?" This chapter looks at the techniques for measuring quality continuously so that we can compare the quality of any software build against any other and thereby objectively assess whether or not it is better. By contrast, many projects only measure quality with functional tests, and only carry this out late in the project. This is a serious mistake. We suggest a variety of ways in which the quality status of builds can be established, including testing, automated metrics, automated audits, document generation, and inspection.
Chapter 7, "The Micro Step: Design and Implement," is a look at the work of software development day by day. It considers the tasks in taking one or more requirements and adding that functionality to the current build. We also look at the techniques you will employ using Together ControlCenter for designing the functional tests, the user interaction, the object interaction and persistent data, as well as documenting the code and links to the requirements. Elements of Together's Interactive Development Environment (IDE), such as its editors, GUI-builder, debugger and source-code formatter are mentioned.
Chapter 8, "The Macro Step: Architecture," considers the system at a different scale and perspective. Many different definitions of software architecture have been used since the phrase was first coined, reputedly by F. P. Brooks. For us, architecture means the principle divisions of the software and the policies and patterns used to extend the software. In small teams, architecture is not the job of a separate team, and for this reason some teams simply let it evolve. We believe, though, that architecture is too important not to be given separate consideration. In this chapter we look at how the requirements for architecture are derived from non-functional or "level of service" requirements. This chapter also gives us an opportunity to consider how the package diagram, with Together's dependency analysis, is a key weapon in the software architect's armory and how component and deployment diagrams also help.
Chapter 9, "The J2EE Architecture," looks at the specifics of architectures based on Sun's J2EE standard and how Together ControlCenter provides support for this. We also talk briefly about our favorite technique for mapping objects to databases.
Finally, Chapter 10, "Parting Words," gives us an opportunity for a brief reflection on the scope and limitations of this volume.
Throughout the book we'll show you how to get the most out of Together, whatever formal process (or lack of it) you may have adopted. For example, we'll be exploiting Together's open APIan extensive Java-based API that constitutes a comprehensive plug-in architecture for Togetherto develop modules to make development and documentation easier. We hope it will encourage you to explore this powerful and flexible facility in Together ControlCenter. Together is, after all, a development platform rather than a development tool. As teams discover specific tools, patterns, audits, or integrations they need, but which are not yet delivered out of the box, the open API is the obvious place to turn.
And that's it. We're enthusiastic about Together, and we hope it comes across. If you're as enthusiastic by the time you finish the book and have explored its accompanying software, then we'll be happy. If it gives you a different perspective on how to build software in an efficient and quality-focused manner, we will have achieved our goal.
The version of Together referenced in this book is Together ControlCenter, Version 6. In some cases pre-release versions of the software were used, and so the functionality or user interface may differ somewhat in the version that you may be using. Each version of Together ControlCenter sees improvements and enhancements to its user interface, so we have avoided making instructions to navigating its interface too specific, lest those instructions become out of date in a future version of the product.A book like this is never just the work of those whose names appear on the cover. We are immensely grateful for all those who have encouraged us, advised us, argued with us and worked with us over the years, and whose ideas and insights have inspired and honed our own. Our special thanks go out to Alex Aptus, Bruce Anderson, Dave Elton, Dave Astels, Dietrich Charisius, Eric Lefebvre, Gareth Oliver, Jo Perry, John Nicholls, Jon Kern, Karl Frank, Ken Ritchie, Laurence Wilson, Mike Swainston-Rainford, Miroslav Novak, Nick Dalgliesh, Paul Field, Paul Kuzan, Peter Coad, Randy Miller, Richard Pitt, Robert Palomo, Ron Norman, Steve Palmer, Tim Shelley and Tom Lee.We'd especially like to thank Robert Palomo, Bruce Anderson, Richard Pitt and other reviewers for their helpful comments and their improvements to the manuscript and case study software; also our thanks to Paul Petralia and Mary Sudul for their editing and page layout work respectively. We'd like to thank Peter Coad for commissioning us to write this work, and for the inspiration his own writing has provided. Pete has the courage to simplify, which is why his work is so valuable. It is something we have striven, not always successfully, to emulate.Finally we would like to thank our families: Sarah, Amy, Laura and Olivia Carmichael; Sue and Phoebe Haywood. Thank you for your understanding when our minds and diaries are filled with such trivia as improving software development. But thank you most of all for what makes working for living worthwhile-your love.Andy CarmichaelSouthampton, England.Dan HaywoodMaidenhead, England