This paper was written as my final paper for my graduate Process Management course at RIT. The paper is based on a semi-fictional software as a service (software on demand) vendor which develops its own software and maintains its own development processes. This document was to be formatted somewhere between a handbook and a memorandum to the CIO from the Director of Software Engineering. So the document is somewhat technical and not all terms are explained. Because of the target of the paper this document is a mix of theory and everyday practicality. For example, at times we will discuss development methodologies but at other we discuss actual hardware specs that will be out of date almost immediatley.
To simplify things, the company in question is simply CompanyX and their main product is ProductY. This document was originally a Word doc which I had to paste into Notepad and then into WordPress and then manually edit to restore formatting. So please forgive any formatting mistakes as it was a lengthy manual process.
As requested, I have assembled a body of knowledge and best practices in process management with which our organization may promote healthy, productive software development and software development management. Because CompanyX focuses solely on software as a service (i.e. “on demand”) products, many traditional approaches to process and project management do not apply or must be applied with an eye towards modifying their principles to work most aptly within CompanyX’s corporate framework. This document is an attempt to distill the general body of knowledge available in this area and to modify it so that it is applicable and usable within CompanyX.
As you know, CompanyX’s approach to software development has always focused around forward-thinking, progressive hiring practices which aggressively seek to hire the absolute best IT professionals and extremely agile development practices designed to leverage the skill base of our developers most effectively while reducing overhead created in other organizations necessary to manage mediocrity in the environment. Because of these practices it is necessary for any software development discussion within the CompanyX context to take these into consideration and to therefore alter generally accepted industry norms and practices to reflect this.
The basis of this document is not to enact change unnecessarily but to identify potential areas for improvement, to find areas which are weak, missing or poor and to codify practices which are working and should not be changed. This document is very timely as CompanyX is currently undergoing a change from a single-focus medical facilities software firm to a multi-focused firm with a financial software group and internally managed hedge fund.
Current Development Status
To begin our discussion, I would like to look at the current state of the CompanyX development environment to establish a baseline against which we may compare our proposal of best practices and design.
CompanyX has two key datacenters, Chicago (internal and development) and Houston (production and Internet facing / non-VPN applications.) Houston is in the process of being relocated to San Francisco. A third datacenter, dedicated to backup storage and warm recovery needs, is being planned in Philadelphia.
At this time, CompanyX employs only a handful of developers most of whom do not need to directly collaborate as each works in a highly segregated technology niche which creates a situation of relative exclusion simply by its nature. All CompanyX developers work from home offices provided by CompanyX using CompanyX workstations, software, networking hardware and telephones (when appropriate.)
Collaboration is handled through the secured corporate XMPP and eMail servers as well as over VoIP. Remote desktops are visible using the central NX remote desktop acceleration system giving all employees access to one another on the central network. Centralized documentation is handled through the corporate wiki. Source control is handled via Subversion hosted at Peoria Heights – closest to the development servers.
Each home office has a hardware firewall with IPSec VPN acceleration to make networking to CompanyX’s Peoria Heights office transparent and simple. Files can be transferred directly via CIFS or NFS from the central filers.
The main development platform at this time is Visual Studio 2008 using C# 3.0 and ASP.NET 3.5. Visual Basic and VB.NET has been eliminated after years of weaning. SQL Server 2005 and MySQL 4/5 are the key database platforms. PHP is in use for some internal applications.
Given CompanyX’s current small scale this environment has proven to be quite effective, but we must realize that as growth and diversification occurs it will become increasingly necessary to add more structure and communication methods in order to keep projects moving fluidly.
Physical Location Issues
The first issue in development environments that we will address is physical space and location issues for the development teams. Conventional wisdom says that the most practical means of getting maximum performance from developers is to locate them physically near to one another to facilitate rapid and smooth communications. The need for rapid communications, however, is in direct conflict with the need that developers have for quiet and uninterrupted time for development to allow them to efficiently enter and maintain “the zone”.
“…knowledge workers work best by getting into “flow”, also known as being “in the zone”, where they are fully concentrated on their work and fully tuned out of their environment. They lose track of time and produce great stuff through absolute concentration. This is when they get all of their productive work done.” – Joel Spolsky, Joel on Software
In weighing the benefits of high level communications versus the benefits of isolation, flexible time, flexible location, travel, etc. we have determined that the cons of collocated development teams outweigh the benefits in our case. Our approach, at this time, is to continue to pursue a completely work from home environment encouraging extremely flexible schedules allowing developers to maintain “the zone” for longer than usual durations while encouraging extended downtime when “the zone” is not obtainable.
Regardless of the physical juxtaposition of developers a superior development environment is a necessity. Following the recommendations of Joel Spolsky, a recognized expect in physical development environments, we intend to make Herman-Miller Aeron chairs available for all developers along with two 20.1” LCD monitors as a minimal, standard setup. Additional monitors and choice of development platforms (Windows, Mac OSX, Linux, Solaris) are available at the request of the developers. Also, all developers have access to commercial grade Bush desks that can be shipped directly from or picked up from Office Max. These desks have surfaces that are very conducive to heavy mouse usage.
Given that intra-team communications are of the utmost importance but recognizing that we are unable to effectively co-locate most of our staff we must turn to technology to mitigate this problem as far as is possible. This is a challenge faced by all teams even when they have the luxury of physical proximity. We believe that we hold an unapparent advantage in this area because we are acknowledging communications as a shortcoming up front and tackling this issue seriously rather than assuming that communications is handled vis-à-vis natural proximity.
We have a number of tools at our disposal that should function well in the role of providing a solid communications infrastructure for CompanyX. We have many of these tools in place already such as a very successful email system and groupware package that handles basic communications both internally to the company and externally to clients and vendors. The system is secure, reliable, readily accessible both inside and outside of the company, is tested to work well with BlackBerry and iPhone devices, etc. We have team aliases set up to encourage team collaboration in this form.
Our email system also includes a simple document management system which has not proven useful for team collaboration but has shown to be very effective for individual user note-taking and reference information rather than storing information directly in an email format but still allowing this data to be accessed easily through the convenience of the email interface.
Group calendaring is currently proving to be a very useful addition to the eMail and groupware system. Calendars are easily viewed by interested parties and teams can select calendar views to allow them to see the available and busy times of everyone on their team all at once in a single calendar view which is very effective for ad hoc planning and collaboration events.
In addition to asynchronous, archival communications available via eMail, CompanyX focuses heavily on the use of the XMPP/Jabber protocol for instant messaging via the OpenFire Server product. This tool has worked very effectively for us in the past and our plans for the future involve a continued investment in this technology and an advancement in our client-side architecture with which to support communications. Currently we have a Java-based client, Spark by Ignite RealTime, that serves very well for general use. We have begun testing an Adobe Flash based client that is available via a web page that may prove to be very useful for staff who need casual instant messaging access from machines where Spark is not installed. It is our goal to make secure instant messaging access simple and ubiquitous so that staff are available to one another, in real time, whenever possible.
Often staff is required to be live on the instant messaging infrastructure while located in a secured office location that does not allow direct XMPP/S communications to the Internet. For these users we have implemented an SSL Proxy / VPN product, SSL-Explorer, which allows them to access our applications such as eMail and XMPP even under the harshest network conditions.
Recently, CompanyX has moved from pmWiki to DokuWiki and has begun focusing heavily on using the wiki format to support as much internal development documentation as possible. The wiki application is ideal for much of development communication because it is easy for any collaborator to create necessary documents, all contributors can edit their appropriate documents and staff requiring information have a central, searchable repository in which to perform searches for necessary information. Wiki documents are automatically versioned making it trivial to track changes over time and to return to previous versions when necessary.
The wiki format offers many benefits that would be considered non-obvious. DokuWiki allows staff to subscribe to wiki pages in which they have an interest getting automatic change notifications, either via eMail or via RSS/Atom feeds. Through this subscription system the people who are responsible for or dependent upon a particular set of data can find out about changes immediately and react accordingly even if the person modifying the document is unaware of which parties may be affected.
The wiki format is useful for many types of documentation with standards documents, best practices, guidelines, project initiation documents such as the project charter and use cases being among the most effective. The wiki is most useful as a supporting architecture for use cases which are the bread and butter of our design documents at CompanyX. Their text basis and ongoing modifications lend them to be ideal candidates for wiki-ization.
Not all document types lend themselves well to the wiki architecture. Rarely changing documents, end-user printable documentation and UML diagrams, for example, are more appropriately handled through a richer format such as Microsoft Word, Microsoft Visio, Umbrello and OpenOffice.org Writer. To support these file formats a more traditional content management system is recommended. The two products that are leading, at this time, in that area are Microsoft’s SharePoint Server and the Alfresco Content Management System. Both of these products often portal functionality so that they may function as a central starting point for staff when searching for content as well as document repositories with interfaces to traditional file stores. Both may additionally be used as supporting frameworks for module application interfaces.
Voice communications remain very important, especially for detailed technical discussions in real time. In order to support fluid and cost effective voice communications CompanyX will implement an internally hosted Voice over Internet Protocal (VoIP) system. VoIP compares favorably to traditional voice technologies with many advanced features available such as conference calling, call forwarding, follow me service, voice-to-text support, auto-paging, voice to email interfaces, etc. The cost of VoIP is also very low compared to traditional voice technologies. Because VoIP systems are so cost effective, CompanyX has the opportunity to extend the internal corporate voice infrastructure into the homes of all staff providing a seamless environment so that neither internal staff nor external clients can easily tell when a given member of staff are located at home, on the road or in the office. This flexibility and transparency is important in supporting the disparate office environment required by CompanyX.
Unified Communications platforms, such as those made popular by Microsoft, Cisco, Nortel and Avaya, have many advantages over our recommended approach of maintaining and moving forward with separate eMail, IM, VoIP and other communication products. By choosing to work with separate products CompanyX has the benefit of choosing “best of breed” products rather than a single, unified product that does not necessarily need to compete head to head with industry leading products. More importantly, however, Unified Communications or UC fall under a legal grey area, at this time, and may subject loosely regulated communications such as voice to long term scrutiny and archival penalties of communications such as eMail. Because of laws within our primary jurisdiction, the United States, it is important that UC be avoided until legal hassles surrounding its implementation are resolved conclusively.
Development Methodologies are critical underpinnings of software development. CompanyX has always focused on highly iterative and extremely agile development practices focusing heavily on those methods and practices that lend themselves first and foremost towards heavy innovation, rapid adaptation to the market and quick feature development.
Because of our needs and strategies, CompanyX has turned to highly modified Agile methods drawing inspiration from popular Agile methodologies, especially eXtreme Programming (XP) and SCRUM.
CompanyX has adopted the XP use of the “user story” as the primary measurement against which development is performed. Stories are not completely “user” related and can sometimes be “developer” stories to allow simple scheduling for changes not visible to the end user. This can be misleading to XP adherents as stories are designed to be visible to the client but CompanyX’s true end client is, in a fashion, the development organization itself and so we treat stories in an unusual way at times.
In order to support the continuous development processes within CompanyX any standard methodology must be adapted heavily as all major methodologies rely upon release cycles, sprints, etc. CompanyX uses Continuous Sustainable Velocity as its model and does not have scheduled release cycles.
A development team “burn down chart” or backlog system is used to schedule upcoming stories so that priorities are maintained even without a set schedule. The backlog is continuously managed by the development lead and the analyst lead / project manager to keep the stories organized by priority but with continuous reference to their weight (estimated time for completion.) The weight may be used to determine that a small feature or minor change might need to be popped to the top of the backlog stack because it represents better story velocity.
The goal within CompanyX is to get new features designed, developed, tested and into the hands of our customers as quickly as possible. Because we release all of our software directly to our own servers and deploy internally making the software available as a service to our customers we focus on releasing as features are made available and do not need to announce most new features or changes. We try to roll out changes gradually on a regular basis so that customers always have some new fix, feature or performance enhancement. To our customers our system is very much alive – a system constantly under change and development.
CompanyX’s methodology also follows the XP principle of simplicity and “You Aren’t Gonna Need It” or YAGNI. This principle, which states “Always implement things when you actually need them, never when you just foresee that you need them.”, is important to allow CompanyX to produce software more rapidly while delaying potentially complicated and unneeded work until we are sure that it is required.
CompanyX loosely follows Mehdi Mirakhorli’s Rules of Engagement and Rules of Play for XP which I will list and explain:
- Business People and Developers Work Jointly: CompanyX follows this very closely but, because CompanyX does not have direct customers but is itself the primary customer of the software products which it then hosts as a service, CompanyX produces an industry trained business analyst who represents CompanyX as an entity as well as theoretical business customers. This analyst is available full time as a “business person” with whom the developers can go for clarification, expansion, etc.
- The Highest Priority is Customer Satisfaction: This means primarily that CompanyX is able to host and manage the application and second that CompanyX’s SaaS customers are thrilled with the quality, features and innovation of our services.
- Deliver Working Software Frequenty: In XP, short term timeboxing is used to encourage the delivery of new features and updates on a regular basis. At CompanyX we attempt to roll features and fixes out as quickly as they become available on a continuous basis. Delays are only introduced to this system when multiple changes may be available extremely close to one another or when system changes are considered risky.
- Working Software: Our goal is to produce working software. No other deliverable has the priority of working software.
- Global Awareness: The development team must tune its behavior in order to support the overall needs of the organization and its customers.
- The Team Must Act as an Effective Social Network: Honest communication leading to continuous learning and an emphasis on person-to-person interaction, rather than documentation, empowerment for developers and analysts to do the work that they need to do and to make the decisions necessary to create and innovate and to level the political playing field so that all team members are available to one another and seen as peers.
- Continuous Testing: All code is continuously tested through all reasonable means.
- Clearness and Quality of Code: Code should be clean and elegant resisting duplication and obfuscation. When possible, code should adhere to the precepts of Software as Literature.
- Common Vocabulary: The team, as well as the organization, should share a common vernacular to reduce miscommunication and increase the speed of development.
- Everybody Has the Authority: Everyone on the development team has the authority to do any work, to fix any problem and two or more people are always available who understand any given piece of the system.
In addition to the “Rules” of XP which CompanyX follows almost exactly, there are twelve XP Practices, as set down by Kent Beck, which we follow less closely. The practices which we do not believe are beneficial to us include pair programming and test driven development (at all times.) We feel that these practices are too heavy for CompanyX’s needs.
The XP Practices which CompanyX should be adhering to include: the Planning Game (for CompanyX this is played between the development and analyst leads), Whole Team (business is always available for the developers), continuous integration, refactoring, small releases (CompanyX uses continuous releases), coding standards, collective code ownership, simple design, system metaphor (business and development joint metaphor for system functionality creating a meaningful, shared namespace) and sustainable pace (CompanyX should never approve pushed overtime.)
CompanyX also takes concepts from the SCRUM development methodology but not as heavily as from XP. SCRUM’s focus is more upon agile software management then upon software development itself as is the focus of XP.
SCRUM separates stakeholders into two key groups, pigs who are completely invested in the project and chickens who have an interest in the project but are not necessarily totally committed to it. CompanyX follows this tradition but without the humorous naming conventions.
Within the pig group of stakeholders, SCRUM recognizes three key roles: product owner, ScrumMaster and the Team. The product owner is the voice of the customer and makes sure that the team is working on features and in a direction most beneficial to the business. The ScrumMaster, sometimes called a Facilitator, whose role overtakes the traditional role of the Project Manager but, instead of leading the team or “managing” them the ScrumMaster’s job is to interface with the organizational management and to make sure that outside influences do not get in the way of the team’s ability to work. Finally, the team itself includes the developers, designers and others who perform the “real work.”
CompanyX organizes its teams similarly. The Product Owner role is played by the business analyst assigned to the project who represents the clients or potential clients. Since SCRUM is generally organized around contract software development their role is more aligned with a contract customer while CompanyX’s customers are potential industry customers and so the business analyst’s role is to have a firm grasp on the business of the industry as a whole and to represent them in a way that no one person within the industry may be able to do so themselves. These roles are extremely similar.
The ScrumMaster role is less clearly defined within CompanyX. As CompanyX does not actually use SCRUM we do not require someone to manage the SCRUM process. A manager is selected to act in the Facilitator role, but we use the term Project Manager, whose role it is to keep the team able to work. The Project Manager is not over the team and does not manage the team but manages the project, interfacing with the sponsors, providing data, keeping the team able to keep working by not being bothered by the overhead of politics. Often this role does not require a full time person and this role is either played by an executive sponsor or by the business analyst.
CompanyX’s concept of a team aligns very well with the SCRUM idea of the team being a self-organizing entity that needs no external management. Like SCRUM, CompanyX keeps teams small so intra-team communication is fast and fluid. Additional overhead of formal meetings or project management processes are not necessary.
Much of CompanyX’s concepts of project task organization come from SCRUM. CompanyX’s teams utilize the product backlog and burndown chart systems for providing straightforward and simple project planning, scheduling, etc. As we do not use Sprints themselves the idea of the Sprint Backlog is missing and, as we see it, unnecessary. CompanyX does not require schedule detail of the level that is generally provided by the Sprint backlog and so this is seen as superfluous overhead that should remain out of the picture.
Unlike both XP and SCRUM, CompanyX does not generally encourage customers to become a part of the development process. Ideas and suggestions are solicited from customers at times but these are used to provide vision and guidance but almost never day to day project goals or scheduling. We choose this model because CompanyX delivers a software product that is then purchased by unknown customers. It is the job of the business analyst to understand the needs of the potential customer pool and to anticipate their needs better than they would be able to do.
After eight years of developing our methodologies based upon our studies of the software management and development field and upon our unique software development models and delivery mechanisms we feel that our blend of light management, informal processes and extremely agile methodology is quite effective and does not constitute unnecessary overhead while allowing the team members the ability to communicate and work effectively together.
Core Technology Competencies
Due to CompanyX’s relatively small size it is important that we focus upon a core group of technology and platform competency areas in which we are able to specialize and develop a mature practice in which we are able to cross-train and mentor. The core areas in which we are currently focused include Microsoft’s C# on the .NET platform, Java with Spring and Hibernate on UNIX (both Red Hat Linux and Sun Solaris) and Ruby on Rails. C# and .NET have long been our core platform replacing our previous VBScript/ASP platform based on Microsoft’s venerable DNA model.
C# and Java remain our most critical platforms providing the basis for much of the code which we currently have in production as well as much of the code that currently exists in our pipeline. These two languages and library collections are also, by far, the most well known and well understood platforms by our current development teams who have used these technologies extensively both within CompanyX and without.
In the past year, Ruby and the Rails web application framework have begun to appear as an important additional platform choice for CompanyX. Ruby on Rails is targeted at rapid web application development and often aligns well with highly agile methods like those used at CompanyX. Ruby is well suited to the average database intensive but computational non-intensive applications which we are apt to produce. Ruby’s high level of abstraction and powerful library system is highly advantageous for building complex business applications very quickly.
For financial applications and other high computational performance needs situations the use of C or, less likely, C++ may be critical for small pieces of code requiring extensive tuning. This may not be a correct assumption, however, has the Java platform is extremely performant, especially in numeric computations, and may prove to be more than adequate. Having a C compiler tested and approved would, most likely, be important to ensure that necessary tools are ready before they are needed. Our financial development lead is currently evaluating two compiler packages in order to find an appropriate choice for our platforms. As we are standardized on AMD Opteron and Sun UltraSparc processors only the GCC (AMD Optimized) and the Sun C Compiler (Sparc Optimized) are being considered since both outperform the rather expensive Intel C Compiler for our chosen hardware platforms.
In the future, having access to a readily available C platform may be critical as CompanyX investigates the NVidia GPU-based CUDA technologies for massively parallel floating point computations. The CUDA libraries and drivers necessary to access the NVidia Tesla GPU hardware are currently callable only through C which may necessitate the use of this language.
Under investigation, potentially as production development platforms but more often as learning and exploration opportunities, are a number of languages and platforms. We believe that it is important to constantly seek out new languages, approaches, platforms and techniques so that we are prepared when new challenges arrive. Experimenting in new languages and platforms is beneficial outside of potential production usage as well. Having an opportunity to work in a diverse environment is more exciting for developers and also contributes to growth.
Currently we are experimenting with the use of Clojure (a Scheme-like LISP language that runs on top of the Java Run Time), Groovy (a Ruby-like language that shares syntax with Java and runs on the Java Run Time) and Scheme (a LISP variant.) Highly potential other languages for our use include Scala, Boo, J# and Haskell. By working with new and groundbreaking languages we expand the way in which our developers think helping them to approach problems from a unique perspective and to discover new and different solutions.
Standardized design tools are important in providing consistent communications and documentation from the design process to the development process. CompanyX chooses to focus on a limited set of highly effective design tools that deliver us the maximum of communicative power without unnecessary overhead and the creation of design documents that are unlikely to be used and may rapidly become outdated and ineffectual.
We have chosen to use the wiki format as our preferred mode of design document sharing. Wikis are easy to implement and easy for all team members to access and modify. Collaboration happens fluidly and naturally on a wiki and built in change control mechanisms protect against data loss even when major changes to documentation are made. In the past we have used pmWiki but found the system to be lacking in many respects. After some searching we have, in the past year, migrated to DokuWiki which is much better suited to corporate-style documentation needs and authentication options that allow us to integrate it into our environment effectively through LDAP integration which allows it to authenticate either to our UNIX systems or to Active Directory and Windows.
The wiki format has proven to be very effective from its initial usage at CompanyX and we have begun to use it in many areas including new hire documentation, helpdesk information, application and package instructions for desktop applications, scripts for deskside build and migration tasks, general company data, etc. Our goal is to move as much information, except for financial control data, into the wiki format as the format will allow both because we find this system to be very effective for our small firm but also because by putting more information into a single repository we find that usage rates increase and more employees are able to find more types of data more quickly.
At CompanyX we have chosen the Use Case as our primary form of design documentation. Use Cases are simple for all parties to create, consume and comprehend and provide the most important piece of documentation needed by the development team – a clear understanding of the customer needs. Use Cases are divided into business use cases and system use cases and are designated as each within the wiki. Business use cases come from the business analyst and include business process modeling. System use cases come from the system analyst / architect and involve technical, system process models. Comprehensive Use Case documents are a long term project resource since they reflect the way in which the system will be used or the way in which we believe that it will be used. As such this documentation is much less likely to become outdated than more technical documents are wont to do. In some cases, Use Case information close to a decade old has been beneficial in passing long term project knowledge between project iterations.
Beyond the use of Use Cases, which are often mentioned in conjunction with the Unified Modeling Language (UML) but are not actually a part of the UML specification, CompanyX makes selective use of UML diagrams and models for more technical aspects of the system design process. The graphical nature of UML diagrams makes them naturally more difficult to store, catalogue, search and retrieve. Because no system, of which we are aware, is effective at dealing with a store of UML diagrams we have chosen to use the wiki system again as a UML diagram repository alongside the Use Cases. In some rare cases a UML Use Case Diagram might be attached directly to a Use Case but more likely a separate category of pages is created with carefully organized diagrams arranged as images on the wiki page with appropriate text on the page making the page as searchable as is possible.
While our use of UML is less extensive than Use Cases we find the more technical diagrams such as class diagrams and object diagrams to be the most effective. Other types of diagrams would be used only rarely when a special circumstance called for a high degree of clarification.
UML diagrams are created using the Umbrello module of the KDevelop Developer’s Suite. This package is available on all standard CompanyX Linux desktops. Umbrello was chosen as the best UML modeling tool for our purposes after auditioning several other tools such as ArgoUML, Kivio, StarUML and Poseidon. For some special case needs Microsoft Visio may be required. This is available on all CompanyX standard Windows desktop builds.
In addition to traditional design tools, CompanyX, being completely non-colocated, requires special attention to the needs of team communications in a collaborative development environment. In order to support these forms of communications, which are especially critical during design phases, we have traditionally implemented instant messaging conference rooms and advanced voice telephony functionality that includes extending the organizational voice system to all offices and employee homes. These tools are key for rapid communications but play only to synchronous or near-synchronous communication needs and do not address long-form discussion, discussion tracking or decision archiving.
To support these longer duration needs, we are suggesting that an online forum application be implemented that will allow teams to pose questions and ideas and invoke response and discussion both from the team itself as well as from the community. Many open source forum and bulletin board packages exist to provide this functionality. We have previously auditioned phpBB which is, most likely, the most popular package available in this space. A new product, Vanilla by Lussumo, is also a recommended candidate. Another potential candidate is Beast, a Ruby based forum.
Developers work day in and day out within development environments and because of the percentage of time spent within these tools it is very important that these tools be readily available and appropriate for the task.
All developers should receive moderately high performance workstations which are more than adequate for software writing. Hardware is effectively free in comparison to developer time and effectiveness. The IT team keeps a standard desktop build design for hardware and software designed to be efficient and powerful whilst giving developers the flexibility that they need to be effective.
Because CompanyX has a highly “at home” workforce and due to the fact that our desktops need to be transported regularly it is useful to keep our machines as light and small as is reasonable. The recommendation is to use the Hewlett-Packard Small Form Factor dx5xxx family of 64bit AMD Athlon64 and AMD Phenom based desktops. These machines are available natively with 64bit Windows Vista to support the latest Microsoft-based development environments and are completely compatible with our Novell OpenSUSE 64bit development environment. In some special cases, Apple Mac OSX or Sun SunBlade Solaris desktop may be requested.
CompanyX should make a concerted effort to rapidly replace all aging 32bit systems still in the field with 64bit hardware and software as all deployment platforms have already migrated to 64bit exclusively. Development will be made more simplified if only a single architecture standard is in use.
Many companies choose to use full-fledged multi-processor workstations rather than high end desktops, such as are recommended here. Those are very valuable and should be made available upon request but most developers, we believe, are unlikely to wish to have large pieces of equipment such as these in their home offices and they are unwieldy to transport making them a less advantageous choice compared to the less expensive small form factor (SFF) desktops.
Desktops should be outfitted with very fast, multi-threaded processors and extensive memory allotments to support rapid development, responsive environments and easy, local compiling. Care should be taken if any deviation from the IT department standard hardware is made to ensure that the hardware selected is able to support multiple monitors as many desktop products, including units in the same lines as we use, are unable to do so without the addition of extra hardware which adds expense, support problems, time to deliver and complexity.
In addition to dedicated hardware, we have traditionally offered a remote desktop option for users working from remote locations or who just want access to an internal CompanyX managed Linux desktop sitting inside of the corporate firewall already configured with a large set of tools for development. To enhance future developer flexibility it is recommended that this system be upgraded to newer hardware and software and designed to handle many simultaneous remote users nearly equivalent to the entire development team. This remote desktop service should be upgraded to mirror our most current Linux standard builds to provide consistency between the desktops and the remote access services. Also, currently Windows developers needing remote access to Windows development environments do not have a reasonable option to accomplish this. The proposed solution is to add additional datacenter functionality that will support direct, remote access to developer’s personal workstations through a central accessibility portal.
For Windows development, CompanyX has been quick to adopt the latest in Microsoft VisualStudio platforms moving to each one as it has become available moving from VisualStudio 6 through VS.NET 2002, 2003, 2005 and currently 2008. VisualStudio provides a complete set of Integrated Development Environment (IDE), compiler collection and database design tools. For our standard development package we believe that this set of tools is more appropriate for our chosen C# / .NET platform.
For UNIX development, CompanyX has chosen to use OpenSUSE Linux on the desktop and Red Hat Enterprise Linux on the server providing a nearly seamless set of tools. Linux is our operating system of choice for most development languages including Java and Ruby. Unlike on the Windows platform there is not a single, obvious source for all development needs on UNIX. We have chosen the Eclipse IDE as our corporate standard for Java and Ruby development. Eclipse is available on both Linux and Windows providing additional opportunities for platform independence.
On occasion, C and C++ develop is needed. For this purpose we have decided to provide two options. The first option is C support in Eclipse. This is useful for most C tasks. The second option is the Linux-specific KDevelop IDE which is well suited to development targeting the Linux KDE desktop and the Qt Toolkit.
Continuous Sustainable Velocity
CompanyX has long valued the importance of providing an environment where developers are given the best possible opportunity to innovate, design, create and write great code. We believe that introducing excess overhead, including most deadlines, does not lead to better code or better innovation. We want to be sure to provide our development teams with the best possible tools and environment in which to create while doing as little as possible to hamper their desire to produce. Taking a page from the book of academic procedures, we see all of our development as being a part of “research and development” and encourage any flexibility necessary to encourage this type of behavior.
Continuous Sustainable Velocity, or CSV, is basically the aforementioned ideal. The idea behind CSV is to create an environment and management structure to support a maintainable maximum development velocity without exceeding fatigue thresholds. Traditional projects use deadlines as a means of pushing teams to finish code more rapidly. Often this works but at the expense of slowed development after the push and a higher rate of bugs and errors. This type of management behavior sacrifices long term gains in return for short term benefits and overall produces less code and what code is produced is of lower quality. This is similar to Wall Street firms forfeiting long term financial growth in return for short term stock price inflation. CompanyX does not have to answer to short term gain seeking investors nor does its software delivery model create a need for hard development and release cycles. We are in a unique and enviable position to exploit these advantages.
CSV does not eliminate the potential use of iterations, that is not the intent. The idea behind CSV is to eliminate deadlines which are set by management, marketing, sales, customers, etc. All products are created on a “when it’s ready, it’s ready” basis. Management works with development to determine “release criteria” which determines when certain milestones have been met. Using time estimates from developers as approximate completion dates to provide managers with the ability to create meaningful “drop dead lines” for burndown task lists to mark when an iteration, version or milestone is complete provides us with a scheduling mechanism, but one very different from what most companies use. In this way some amount of scheduling can be achieved without inflicting deadlines upon the development team. Instead, deadlines are inflicted upon management.
Because of the Software as a Service model in use within CompanyX there is no need for a formal product release cycle. Instead software can be released at the end of every iteration. In many cases these releases are very small and changes could go unnoticed by customers. If an important feature or fix is ready before an iteration is complete it can be released to production immediately.
CSV is designed to not encourage overtime but to encourage flexible time. The idea is a sustainable pace but with the understanding every day cannot be the same and that without variety there is no way obtain a sustainable pace. Employees are encouraged to work long hours when they want and to not work when they do not want. CompanyX’s goal is to allow each employee to work in the manner in which they feel most productive. For some this means eight hours per day, five days per week during normal, daylight hours. For others this might mean working a constantly changing schedule or working in huge bursts of creativity followed by days of relaxing and not coding. We recognize that people are unique and that everyone works differently.
General Development Best Practices
There are many industry accepted best practices for all types of software development. Among these practices are code review, source control, unit testing, test driven development (TDD), continuous integration (CI) and bug tracking. These are the primary practices which we will be targeting as most important to CompanyX at this stage and which, we believe, can be implemented and enforced quickly.
Code review is an important part of any organization’s attempt to improve code quality at a line by line level as well as a way to create organizational integration by disseminating knowledge, technique and style between team members. By having different team members regularly reading each others code it gives a perfect opportunity for learning, instruction and communications. This is especially true for younger and more junior developers reading through the code of their seniors and mentors.
Code reviews have the obvious benefit of error reduction. The idea here is not that code review will reduce syntax and compiler errors as those will generally be caught by automated tools (except for Python or Ruby where new code paths may be tested in production and not in testing) but that having a second set of eyes on the code will allow for logic mistakes to be caught or give an opportunity for someone to suggest an alternate approach that may be cleaner or more efficient. Perhaps even more importantly, if a developer knows that she is going to undergo a code review she is more likely to make the code easier to read which naturally will also make it easier to support later.
Source Control is an absolute necessity for any development environment. Even teams of just a single developer should be using source control as it is so critical in maintaining records of changes and providing an ability to go back to previous source versions. The larger the team, however, the more critical the role played by source control. Versioning is important but the ability to merge changes from several developers and to manage branch and trunk development is very important. CompanyX has chosen to use the very standard Subversion source control system. Subversion is open source, free and very widely used in the industry. It has become the de facto standard over the previously popular Concurrent Versions System (CVS) which it was designed to replace.
Unit Testing and Test Driven Development
Unit Testing and Test Driven Development often go together as a pair of practices. Unit Testing, often handled through a unit testing framework such as nUnit for C#, JUnit for Java and Test::Unit for Ruby, is a style of code testing that is done by testing the smallest possible piece of the code infrastructure, often the method or function. Unit tests are a very important part of any standard refactoring procedure. By writing the original code and testing against unit tests we have, in theory, proven that the atomic unit of the code is acting correctly. There is still a possibility of error in both the mainline code as well as the unit test but this is far less likely and there is still the very real possibility of architectural or design mistakes but a great number of errors can be caught and eliminated early through the use of unit tests. Unit tests, at a very minimum, force the developer to write the same piece of code in two different ways (a “positive” in the form of mainline code and a “negative” in the form of the expected results) which can have the benefit of forcing the developer to think differently about the problem and may result, in some cases, in faster and more creative problem solving.
The nature of unit tests makes them highly automatable allowing you to use them regularly, such as with each code check in, to determine that new code changes do not break previously functioning code or to verify that refactorings of existing methods continue to work as expected. Unit testing increases the level of confidence in the code and makes integration much simpler.
Test Driven Development uses the principles of unit testing to suggest that instead of writing a method and then writing a unit test to verify the method’s validity that you can instead write a test, which will initially fail, that looks for the expected output of the to-be-written method. Then the method can be written against the test until the test validates as successful. The key benefit generally touted by proponents of this approach is that writing more tests, which TDD causes by its very nature, causes software development to be more effective. Bug levels do necessarily fall compared to test-behind unit testing but productivity has been measured to be higher when tests are written first. It is believed that this is caused by the thought processes and mental approaches to problems taken by the developers when they are writing the tests.
For CompanyX, TDD may be, at least at this time, rather too extreme for us to begin to implement but the use of unit testing, across the board on both new and old projects is very important. Once unit testing is implemented and accepted then experimenting with TDD can possibly follow.
Continuous integration builds on the benefits of unit testing and source control. The concept behind continuous integration is that each time code is checked in to the source control system, which should be a minimum of daily, that a full, automated build process is run creating, in theory, a completely working build of the package and then run the complete unit testing suite against it to verify that integrity has been retained. Should any test or the build itself fail, the package must be corrected immediately to ensure that there is a working codebase for work to continue the next day.
CompanyX has not yet utilized continuous integration in our practices but this practice should be rolled out in conjunction with comprehensive unit testing frameworks. Continuous integration most definitely benefits from having unit tests already in place but we are able to begin before unit tests are actually done allowing momentum to be achieved. Popular continuous integration packages include Cruise Control, Cruise and Bitten. We are assuming the use of Bitten as it is integrated with Trac, which we will discuss below.
Bug and Feature tracking involves a centralized error management system that provides a centralized location for recording bug occurrences, status, priority, etc. Bug reports may be produced internally by developers or testers or externally by clients. By having a central location for bug data managers are provided an opportunity to perform triage on the bugs and can provide developers with a clearly prioritized “to do” list of bugs. This is important in keeping projects moving at maximum velocity as well as for minimizing impact to clients if bugs are appearing in production systems.
Bug Tracking systems are important for any number of reasons arising from the need to know where bugs are, when they have been fixed, by whom and how. CompanyX has, in the past, auditioned products like Bugzilla and FogBugz for this very purpose but was very disappointed in the ability of these products to meet our rather simplistic needs. We have decided to use the Trac product from Edgewall Software. Trac is a popular bug tracking system that has an integrated continuous integration environment feature, Bitten, that can be implemented alongside of the tracking product. Trac is already designed to integrate with a Subversion repository making it a very useful choice for CompanyX.
Project Selection and Prioritization
As CompanyX is a small firm with extremely limited development resources it is very important that we choose projects carefully and prioritize them as best as possible in order to ensure that projects with the greatest potential for success are chosen over those that are less likely to have large returns, have a very high cost to implement or carry a high risk of failure.
CompanyX’s model of software as a service means that our attention is always drawn, first and foremost, to maintaining the quality and integrity of our running applications. Unlike traditional software companies for whom software in the field seldom represents a financial loss if new bugs are found but may actually realize a financial gain through support contracts CompanyX has a vested interest in keeping running applications running as smoothly, efficiently and effectively as is possible. Unpatched, known bugs are very evident to our customers and may represent potential security or performance flaws that will impact our own infrastructure and support costs. Running applications always take full precedence over any forward development project.
New, greenfield projects are started infrequently within CompanyX and choosing an appropriate project is an important decision that affects the company, as a whole, for any single project. For a small firm like CompanyX which is profitable and maintains a relatively even level of developer and engineer resources versus workload in the pipeline possibly the most important selection criteria for a new project is availability of resources.
In most cases, at CompanyX, developer and engineering resources are not interchangeable commodities. Developers and especially analysts are likely to be domain experts in a business area and projects are often selected based on the availability of these skill sets. If new developer or analytical resources become available suddenly a new project may be initiated with very little planning in order to utilize these resources as quickly as possible. In this way we are very opportunistic in our project selection process. Cost / Benefit and other factors are generally subsidiary to resource availability. This is not only because of the importance of resource availability to us but also because resource availability is an easy to determine factor while cost, benefit, MOV, ROI, etc. are extremely subjective and prone to very high rates of miscalculation and misjudgment. Given the availability of a straightforward deciding factor, we believe that we will obtain best results by leaning heaving to the weighting of the known rather than the unknown.
CompanyX is very small and project prioritization is not a significant factor. Projects are generally started when resources become available. There are, of course, a number of projects that are internal or infrastructure related and can utilize any number of different resources and so these do require prioritization, but, by and large, project prioritization is handled naturally but the sudden availability of a market opportunity or availability of resources which are a good match for a particular project.
Prioritization of these projects is handled by executive management and is usually determined by expected ROI. Internal projects tend to be of a moderately similar size and complexity so valuations are relatively straightforward and manpower reduction and competitive advantage are the primary ROI determinations.
Team and Project Organization
CompanyX, being a small firm, needs very little and relatively informal team and project organization. Each organizational unit, being relatively autonomous, generally has a single development team assigned to it. This segregation is intentioned to create an environment where each team feels empowered to self-organize as needed and as is appropriate for the nature of their project. This lack of coordination also provides a more nature freedom for teams to choose architectures and platforms that most fit their needs rather than blindly following trends and directions forged by older teams and successful projects.
Steve Yegge of Google has stated that Google’s availability of great platforms can be a real boon for internal teams looking for resources but that it also squashes innovation by pressuring teams to implement large initial overhead systems as well as square-pegging technologies to work on Google’s existing platforms (i.e. BigTable, etc.) whether or not they are appropriate or advantageous just to justify the former investment in the technology. This justification is clearly one that exists only on paper because if a platform must be used even when it hurts the project then it has actually introduced a negative ROI rather than a positive one.
Because CompanyX does not size teams based on a need to meet a schedule or deadline teams are chosen based on team member interaction, appropriateness for the team, technical needs, etc. This team sizing method is designed around obtaining the maximum performance from each developer rather than picking an arbitrary speed for development and sizing a team until it is able to obtain that speed. This sometimes costs us the ability to get to market as rapidly as possibly in favor of overall per-person productivity. This impact is not as dramatic as it may seem, however, as innovation and high-performance teams are able to move quickly and keep pace with much larger teams that are built without regard for their team dynamic, communications, style and compatibility.
Typically within CompanyX teams range from two to five people. We have found that smaller, more closely knit teams tend to obtain greater productivity per developer and that ideally teams should be in the two to three person range when possible.
Teams are only-lightly managed with self-organization as a high priority. Obviously some management and direction is needed from outside of purely self-organization. Direction is generally established by the business analyst assigned to each business area. The analyst acts much like a manager as they are involved in deciding strategy, task priority, direction, etc. The business analyst is not a personnel manager and no staff report to them.
Management is handled directly from the Director of Information Services. This is the only dedicated management role in the CompanyX IT organization. All IT staff report up to the Dir. of IS. CompanyX is very fortunate to have the ability to avoid the bulk of politics that occur in most organizations simply because of our size.
Virtual staff is import at CompanyX as it is in almost any organization with more than a single team. Because CompanyX is small, though, virtual staff is still a relatively minor consideration. Technology expertise may be shared across teams but this tends to be quite casual. The CompanyX wiki includes information including skills sets and areas of interest of employees so searching for skill areas is possible.
At CompanyX, it is the goal of our managers to make it possible for employees to work. The idea is not that we have hired employees who need to be managed but simply that they need direction from time to time and, more often, need someone who buffers them from the non-development needs of the organization. It is the role of the business analyst and manager, in this case the Director of IS, to eliminate distractions, unnecessary paperwork or other busywork and to provide a politics-free environment as conducive to working as possible. CompanyX’s policy is that the manager is less in control of the team as a resource of the team. We believe that if we hire the right staff and provide them the ability to do great work that they will so it is very important to keep out of their way and not stop them from being productive.
Project Management in the Continuous Sustainable Velocity Organization
In a traditional software development organization, project management is a rather heavyweight process which is heavily involved in the day to day work of the development teams as well as involved in the production of a large number of project artifacts stemming, generally, from a need to report project progress, status, details, etc. to upper management. CompanyX does not operate with a need to produce project artifacts for upper management.
CompanyX management has the ability, and the responsibility, to pull project status by checking the project backlog and burndown lists. In this way status is able to be monitored without needing the team to provide details back to management. The monitoring is passive, not active. At no point should the team stop doing actual, profitable work in order to produce paperwork for internal consumption. Not only does this take time away from the development team and slow progress but it also would require management to spend their time reading status reports rather than being productive themselves.
Because we do not market products before they release and do not have deadlines to meet in delivering products to our customers the needs of our organization are much simplified over more traditional software firms. This gives us an additional advantage in that we can spend more time innovating and producing software at lower cost making it easy for us to outpace our competitors to market.
Some pieces of the traditional project manager role are played by the business analyst. Important decisions such as when a feature is complete or what level of functionality is acceptable are determined by the business analyst.
The PMBOK specifies many tasks which are a part of the project manager’s role but a large majority of these are simply assumed by the creators of the PMBOK and are not actually necessary. At CompanyX, our model is continuous development so we do not “time box” development work into a project that has a definite end date. We cannot budget by release because we do not know when a release will be completed or what it will contain.
Instead of traditional budgetary means, CompanyX establishes an ongoing development budget that is defined in a cost/time manner, such as $65,000/month. This cost includes the salaries of the project team for that duration, an estimate of the months cost of electric, HVAC, Internet access, software, hardware, networking, telephones, chairs, desks, etc. Using this number, which is extremely accurate and varies in predictable ways, project budgetary planning can be carried out based upon the needs for the project to continuously outperform its cost by way of its revenue.
Instead of determining a project’s total cost up front we determine the ongoing cost of a project and determine if we are prepared to support the ongoing cost of development and support. This still requires budgeting and financial planning but the manner in which it is done is far more straightforward and intuitive. It is also far more honest than planning done in many organizations as the ongoing support cost is both difficult to fudge (if resources are assigned, cost is assigned) and total project scope is not a part of the picture. Scope creep, which is controlled by the business analyst, will affect development duration but will not impact the budgetary process.
There is no cause to make early determination as to scope before the project has begun. In order to determine a project budget a normal firm will need to decide upon scope, features, architecture, etc. before the project budget has been approved and, therefore, before the project itself has begun. CompanyX avoids the costly and error-prone process of making scope decisions when far too little is known about the project and, instead, determine a velocity desired and/or budgetary need that can be withstood and allow the business analyst along with the development team to determine feature priority and to decide when the product has enough features completed to allow it to move from purely development into production.
One of the, if not the, most important endeavor undertaken by the staff at CompanyX, or arguably any company, is the accumulation of staff. Because CompanyX focuses less on processes, procedures and documentation and more on team effectiveness, innovation and employee satisfaction we are far more affected by the quality of the staff that we hire. This greater employee dynamic can be seen as a benefit or a deficit. If we hire great staff then we empower them to be creative and productive. If we hire poor staff we empower them to do nothing.
CompanyX has always been and must remain very aggressive in its stance on hiring and retaining the best IT talent in all IT areas including engineering, development, administration, architecture, support, etc. We have found that by the very nature of our management practices and our use of continuous sustainable velocity that we have naturally managed to maintain staff once they have been acquired. Most ambitious and talented IT professionals, especially developers and designers, are thrilled to find an environment so dedicated to respecting their professionalism, providing a framework for their own learning, growth and production, peers that are passionate about their field and the projects on which they work and a lack of red tape separating them from the work that they wish to do.
Far more difficult than retaining great talent is locating and identifying that talent in the first place. At CompanyX it is very important that we find not only great skill and experience but also great drive and passion. CompanyX is a company focused on innovation and to achieve this we require staff that will constantly seek out and discover new avenues for exploration.
Our first means of seeking out great candidates is to simply not seek candidates. CompanyX has always hired exclusively through people seeking out an opportunity to work with CompanyX and through word of mouth. This drastically reduces the chances of hiring someone who is desperate for work rather than seeking a wonderful development environment. We do not want staff that are simply happy to have a job which will pay the bills but want staff who are actually excited about the work that they get to do and would be very unlikely to leave the company because almost no other company offers the exciting, driven environment that CompanyX offers.
When in the hiring and interviewing process for new candidates it is very important that we do not stress specific technical skills. Any candidate may have a select number of highly technical skills for which we test which proves nothing. Technical testing also requires us to test a candidate against a skill set for which we assume knowledge will be needed for the job. This is a short-sighted methodology based around an immediate need and not the long term growth of the company. No technology is so complex that we must require that a candidate bring that existing knowledge with her.
Far more important than a candidate possessing an immediate technical skill is that the candidate has a set of technical skills – any set. What CompanyX actually cares about is the ability for a candidate to grow and adapt. The technical landscape is constantly changing and the skills useful today will not be useful tomorrow. What we are looking for is a candidate who will embrace tomorrow’s skills and learn them easily. Otherwise an investment in an employee is an investment in stagnation. If a candidate earns a job because of a single skill set then that candidate is also more likely to leave later when we move forward to look for “comfortable” work at a firm using legacy technologies.
Even within a single technology for which we would choose to interview, Java for example, there are a range of specific technical skills necessary to be useful on day one and expecting any particular candidate to use the exact same set of technologies and approaches which we are currently using would limit the potential candidate pool dramatically while likely eliminating candidates who could pick up the skill in a weekend and outperform most candidates in a week or two.
For example, within Java candidates may be used to platforms such as Weblogic, Websphere, JRun, JBoss, Spring, Tomcat or WebObjects. They may work with different databases like Sybase, SQL Server, MySQL, PostgreSQL, Firebird, SQLite, BDB, Oracle, Informix, DB2, Derby, etc. ORM packages like Hibernate, iBatis and others. They may use EJB 2, EJB 3 or POJO approaches. They may run on Windows, Mac OSX or UNIX. Perhaps they use Swing to make desktop applications or only write component services. The variation is far too wide to expect that the right fit for CompanyX has been working with the same technology stack as us. If we do find a candidate working with the exact same stack then we risk that that candidate won’t be bringing in fresh approaches and ideas which is some of the greatest value available from new candidates.
Along with technical prowess we also are very concerned with finding candidates who are extremely passionate about technology and are driven not just from a career perspective but also from a desire to excel, to create, to grow and learn and to make, manage or design amazing software products. A candidate with drive but little experience is more valuable than one with much experience but that is not concerned with future personal growth. Working at CompanyX is a lifestyle of technical excellence.
Along with the standard hiring process we must also consider the opportunities that lay in an intern or co-op program. Interns may come from any background including current technical IT work, academia, high school, etc. In an intern we look primarily for drive. Drive is most recognizable in intern candidates of any age who are willing to work on their own projects, at home, in their own time and have been doing so long before becoming interested in working with CompanyX. This same drive is sought after in any candidate, not just one at an intern level, but unlike more senior candidates who have an opportunity to differentiate themselves through career experience, skill set areas and previous projects, interns are viewed almost exclusively through their “at home” experiences.
Recognizing two important facts about academic backgrounds is also important in picking candidates. The first is that many of the best candidates will not have attended college or university because they were too busy already being in the field. The second is that if we desire to add academic experience to a candidate we can easily do that after they are hired; there is no reason to require this before the hire process begins. A tertiary issue also exists that the majority of collegiate IT and CS programs are both far too easy and often avoid core topics that would be expected, and in so doing lessen the potential value in judging a candidate by their academic background even when it appears to be pertinent. Because of these two facts it is CompanyX’s policy to disregard academic experience except in cases where the candidate has undertaken the process of obtaining a degree after having entered and while remaining active in the field. Using university studies as an excuse for absenteeism from the field is not counted as positive experience time but as a resume gap.
Outsourcing is an important topic for companies like CompanyX. Because we have no centralized, physical location we are well suited to both outsourcing and offshoring. We will not, as a company, consider outsourcing labor to another software firm because we both believe that other companies are highly unlikely to be as competitive as we are leading to only higher costs and lower output and also because our value is in creating software and managing IT solutions ourselves. It is these key tenants on which CompanyX builds its reputation and creates its value.
Using labor external to the United States, however, as employees is another matter. CompanyX is officially nationality and locale agnostic. We have no reason not to hire from international locations. Even timezone issues are of little importance as our current staff regularly works floating hours and often works weekends or at night.
Quality Assurance and Testing
At this time, CompanyX does not utilize a dedicate QA and/or testing team for our software. All of our testing is done by the developers and analysts and is done using informal methods. This procedure is far from being ideal and leaves a lot of room for critical improvement. This is an area in which CompanyX should focus heavily in developing a team and process for standardized testing.
The first step in preparing for formal testing is the creation of formal, separate testing and production environments. At this time CompanyX has development environments running on desktops throughout the company and all testing is done directly at that level. We are currently in the process of building out a dedicated testing datacenter which will be used to run pre-production code in a live setting for serious integration and testing work. We currently have a Solaris testing environment built and a Windows 2008 environment is already in process.
After we have a dedicated testing environment, our next step is to begin to build a dedicated testing team. We will need professional testers with testing experience and knowledge who can help us to build a testing practice in the same manner in which we have built a development practice. Implementing both white box and black box testing, testing automation beyond unit testing, interface testing, etc.
We have been fortunate that our small, tightly integrated user base thus far has been resilient to bugs as almost all of them clamor to be involved in early testing but this will change. Our development methodology tends to introduce a low number of bugs, but we should be striving to lower that number as far as possible. As our products are used by a larger and larger set of clients this will become more and more important.
Mentoring is an important element in any professional development environment. Mentoring adds organizational value for junior staff members as it gives them an opportunity to learn and grow with direct access to senior staff members with real world experience and value unobtainable through other avenues.
Mentoring is also important because it provides an opportunity for knowledge to flow back through the organization. There are few mechanisms through which experience both personal and project or technical get disseminated, in a meaningful way, through an organization and one of the most powerful means of doing so is through mentoring.
The process of mentoring also helps to build stronger inter-personal ties within the company. People on different teams or in different technical areas can use mentoring relationships to build networking ties and to become more acquainted with the organization and its resources. Mentoring is significant in cementing corporate culture when new staff come on board.
Project Lifecycle Issues under Continuous Sustainable Velocity and Software as a Service Models
In a CSV and SaaS organization we face particular challenges less common in more traditional software companies in determining exactly what constitutes a project lifecycle and how that relates to its team and team members. When software is delivered as a service it tends to take on a wholly different type of life than does software that is boxed and shipped.
Traditional box and ship software, or other types of software typically delivered in a “versioned” way, has a natural project lifecycle associated with each new version of the software. Often teams will be known by their version such as the Word 2003 and Word 2007 teams at Microsoft. This allows for projects to have a definite initiation point and a clear, or mostly clear, finishing point. Some amount of ongoing support and patching generally continues but is minor in contrast to the project itself.
When working with Software as a Service and especially when doing so without the benefit of regular iterations there is no clear division between software development and its ongoing support. For example, ProductY, our flagship product, ran as a project for approximately eight years. During this time it went through initial development, ongoing support, new features, performance enhancements, stability improvements, underlying platform migration and more.
After eight years the ProductY codebase was replaced by the ProductY 2 codebase. This new codebase we expect to go through a similar lifecycle but lasting approximately twelve years. We anticipate the retirement of this codebase by 2020. In this case, the team responsible for both codebases is the same. Not one person changed between the original ProductY and ProductY 2. From the perspective of the team, ProductY 2 is simply a natural continuation of the original ProductY and part of the lifecycle involved the migration of clients from one system to the other.
When applications go through lifecycles of this nature and on a timescale of this magnitude it is exceptionally difficult to differentiate between different phases of the project. Is initial development until beta one project and then ongoing support another? Sometimes though there is far more code developed after beta than before. Are the two ProductYs separate projects? Most likely that is true. That is the most appropriate project divider that we have identified. Still this is a very long time for the concept of a project to survive.
Because of these issues, CompanyX is faced with almost not even having the ability to identify teams by project at all. This is difficult as the thought processes of developers, and humans in general, invariably brings them to the project. At this time we have not determined the correct answer to this problem but we do have an approach that is, thus far, working while still being a work in progress.
Instead of organizing by project, we have been organizing our teams by application or feature that is a long-term, externally identifiable entity. Therefore, to us, ProductY and ProductY 2 are the same entity as they are a single product to our customers. A separate hazardous waste product, while similar and sharing the business unit, would be a separate identifiable entity and would be a separate “project” internally to CompanyX. This approach, at this time, continues to function but also continues to evolve.
The overall lifecycle of one of these “projects” is very fluid. Project initiation is very similar to a traditional project approach used by any software firm. Initial requirements gathering and basic design is done and then code development commences. Once a potentially useful level of development is deemed to be complete by the business analyst and testing is complete the application is released into production.
Once an application is in production there is no guarantee that it will be used by customers. At this point we generate a basic sales and marketing framework such as a web site, support contacts, etc. The appropriate marketing manager determines if the application is marketable in its current state. If so then marketing begins to attempt to garner interest in the product. If not, application development continues as before attempting to complete more features and increase functionality. The development process is not driven by marketing except in rare circumstance nor is marketing driven by development. Customers seeking our products always have a means to find them even when marketing does not feel that the product is feature-complete enough to justify marketing expenditures but new customers are unlikely to be located until marketing sees the product as more mature.
Regardless of the involvement of the marketing team, development continues on in the same manner. The purpose of the initial lean release is not necessarily to get early adopters but to give customers who are anxious to use our products every opportunity to do so, to get earlier real-world feedback and to put the stake in the ground as to the launch of the codebase.
Throughout the lifecycle of the code development continues. In some cases development may slow as the product ages and becomes more mature. There is definitely a need for more development while ramping up a product than in ongoing support. However, unlike 37 Signals who run a similar model, we are not targeting lean software as our primary market but extremely powerful and robust software with powerful business modeling potential and this gives us more opportunity to mold our products long after their initial release into the market.
In addition to feature growth, old code bases will often see refactoring, bug fixes, performance enhancements, library or platform upgrades, interface updates, etc. which supply a large amount of continuing development work long after initial design and release pieces have been completed. At some point, as we saw with the ProductY product line, a new codebase will likely be generated and at that point development begins anew with a blend of fresh approaches and new designs along with new technologies while drawing from former knowledge and experience.
In the case of the ProductY, the initial system was a web-based Microsoft DNA architecture application running on VBScript/ASP in the Windows NT4 and SQL Server 7 era. It relied upon Internet Explorer 4 as its client side platform and used a small VB6 client component to handle hardware integration. ProductY 2, while providing identical functionality to the customer, does so using C#/ASP.NET on Windows 2003/2008 and MySQL with a single, non-web based client side C# application that handles user interface and hardware integration concerns along with local caching with a client side database. The two architectures and approaches are very different but the functionality is the same.
Packaging and Deployment
CompanyX has a fairly unique need for packaging and deployment of its software. As all of CompanyX’s software is deployed internally to our own platforms and our own systems we have an advantage over most firms in that we have little need for broad platform compatibility testing nor do we necessarily have a need to package our applications for easy deployment.
It is tempting, given the simplicity with which we can have the development teams perform their own deployments through a manual process, to completely forego the entire process of performing application packaging. This makes sense in many ways. It does, however, present a few problems.
The first problem that a lack of packaging creates is an over-dependence upon the development teams. The developers themselves should not be required to be directly involved in the release of a new package, at least not with the physical deployment of the package. By having a package created it allows the system administrators to perform application installations easily and uniformly. This is not necessarily advantageous on a day to day basis but when there is an emergency and a server must be rebuilt quickly having the applications ready, in different versions, for the admins to apply to the environment without developer intervention is very important.
The second problem that arises without a packaging process is that software tends to become extremely dependent upon a unique build and deployment environment. Shares to the environment from a system administration and engineering perspective may inadvertently break packages.
In a Red Hat Linux environment, for example, a customer deployed package may have a library dependency unknown to the system administrator. That library may be updated, moved or removed, breaking the software quite accidentally. If that same software were to be properly packaged as an RPM (Red Hat Package Manager) package then the RPM system itself would check dependencies system-wide and block the removal or changes to other packages needed by the application software. When installing the application software a tool such as YUM (YellowDog Update Manager) would also be able to satisfy dependencies automatically rather than sending the system administrator off to find packages supplying necessary libraries.
For a company like CompanyX, our packaging needs are very simple. Packaging formats such as RPM for Linux (Red Hat and OpenSUSE) and DStream for Solaris (aka Solaris Packages) provide important failsafe mechanisms to make installing fast and simple while protecting the software from inadvertent damage. They also make version control and system management faster and easier. More advanced packing, such as graphical installers, would be overkill in this environment and would, in actuality, make the deployment process more cumbersome considering that all potential deployers of the software are skilled IT professionals and not end-users and customers.
Earlier I spoke about the availability of project backlogs and burndown lists as important tools for management to keep an eye on the status of projects without needing to interfere with the project team itself or to send agents to collect data on the team. In many cases this is not enough information for management or for the rest of the company.
Currently CompanyX works on the basis of the basic backlog and burndown lists along with word of mouth, wiki posts, email distribution lists and a bidaily status and design call in the late evenings. These tools, however, are not scalable and do not provide adequate reporting to parties outside of the direct management chain. More information is needed but it is critically important to keep management from becoming directly involved and interfering with the teams and the projects.
The first status report which should be created for management is the build report from the continuous integration system, mentioned above. Management should get a report, generated every time that the system is run, that details the status of tests passing and failing. This allows detailed progress information to be noted without requiring any additional input from the team itself. This is a good utilization of best practice processes that will already be being put into place.
The second important reporting tool, important for more than just management but for the entire organization, is developer, analyst and team blogs. Having a project/team blog that is updated by the business analyst and/or development lead at least once a week gives insight into the projects for other teams, interested parties and, of course, management. By providing all employees with their own internal blogs there is a system for personal status reporting and organizational information gathering. Tools that may work well here include SharePoint, Alfresco, WordPress and Drupal. Anyone in the organization would have the ability to subscribe to an RSS feed or to visit a website to find the latest information from any person or project.
In additional to traditional blogs which are great for technical posts and project status updates, the more recent concept of the microblog, made popular by Twitter, may also be beneficial to the organization. Microblog posts are generally limited to 144 characters per post, just enough for a sentence or two. This format could be useful to allow developers to post their thoughts, locations, status, needs, etc. throughout the day. CompanyX has even written our own client for making these types of post easily from the command line while working in UNIX systems which is extremely handy for developers to effortlessly make quick statements to the company. For internal use a microblogging platform server such as Laconi.ca would be appropriate, not a public service like Twitter or Identi.ca.
As an example of how microblogging could impact development, a developer might post “Stumped on a C# RegExp.” Someone with extensive C# regular expression experience might see the post and instant message that developer offering some assistance.
Internal Training and the Development of Practice
The idea of Practice within the IT organization is a critical one. By Practice we are referring to the idea of approaching a discipline as a true professional, perhaps even more appropriately, as an amateur in the classical sense. The idea of Practice is, unfortunately, often a nebulous one, but one that is key to the CompanyX approach to software development as a whole.
Practice refers to the holistic approach to the technical discipline whether that discipline is Java programming, Linux system administration, Windows desktop management, project management, business analysis or software architecture. This means that we must look at the individual technical area as an area in which we can learn or develop techniques, strategies, best practices, etc.
Instead of approaching each technical area as just another tool in the tool belt, we look to encourage serious study of the technical areas which may include reading books, studying in a collegiate setting, joining user groups, attending vendor seminars and events, becoming active in online communities, writing and publishing papers, independent research and more.
Internally we support general educational opportunities through support of collegiate academic programs, unlimited book budgets and the Borders Corporate Discount program, internal training, mentoring programs and organized trips to vendor events, seminars and support groups. Most importantly, we believe, it is important that we promote a culture of education and not merely offer opportunities but make learning an important part of everyday life within the organization. We may also want to pursue projects such as having a reading group which meets to discuss readings in important industry publications or discusses seminar talks and panel discussions.
Software Development Best Practices and Standards
Even in an organization such as ours it is important to retain practices and standards throughout the organization. Standards apply in many areas. Applying standards, when possible, is an important factor in making code readable and rapidly adaptable when working with multiple developers.
Two areas in which coding standards need to be applied, across the board, are in naming conventions and code style standards. Both of these areas fall outside of technical areas of expertise and contribute to communications.
For a firm the size of CompanyX, developing internal style and convention guides would be extraneous overhead and unlikely beneficial to the environment. Instead, we have decided to standardize upon the widely accepted Cambridge Style standards. These are well known standards used not only in business but often in academia and widely used in professional publications.
Using these standards assures us that potential candidates are, at the very least, familiar with the standards even if they have not used them directly, will recognize the patterns easily and will write in a common style for the industry when publishing as well. By using such common styles CompanyX can leverage the benefits of style and standard without unnecessary overhead and expense.
Style standards apply both to the “beautification” of code including things like the appropriate use of white space, placement of braces, use of multiple elements within a single line, etc. but also to naming conventions for variables, methods, classes, objects, etc. Through the proper and standard use of such styles and standards programs become more readable, portable, extendable and maintainable. The ultimate purpose is, we believe, to support the concept of Literate Programming as best stated by the legendary Donald Knuth:
“I believe that the time is ripe for significantly better documentation of programs, and that we can best achieve this by considering programs to be works of literature. Hence, my title: ‘Literate Programming.’
Let us change our traditional attitude to the construction of programs: Instead of imagining that our main task is to instruct a computer what to do, let us concentrate rather on explaining to human beings what we want a computer to do.
The practitioner of literate programming can be regarded as an essayist, whose main concern is with exposition and excellence of style. Such an author, with thesaurus in hand, chooses the names of variables carefully and explains what each variable means. He or she strives for a program that is comprehensible because its concepts have been introduced in an order that is best for human understanding, using a mixture of formal and informal methods that reinforce each other.”
Process Management Standards
Many software firms use process management standards such as the Capability Maturity Model Integration and ISO standards to control, improve or monitor processes. No formal process improvement program or strategy is currently in use within CompanyX.
Capability Maturity Model Integration
The Software Engineering Institute at Carnegie Mellon developed the Capability Maturity Model Integration system with the idea that this model would be utilized by organizations seeking to build a framework of continuous process improvement. The CMMI is a set of guidelines as to what processes should be implemented but not how those processes should be implemented leaving the actual implementation up to the individual organization. This is good in that it allows for a great deal of flexibility when it comes to the actual usage of this system with a unique company.
The ideas behind CMMI are almost exclusively targeted at very large, non-IT organizations. These ideas are not necessarily bad but assume a large amount of process overhead which is not present in many companies. Introducing process where it is not needed just for the sake of managing and improving said process is unlikely to be an effective use of available resources. Much of CompanyX’s management approach has traditionally been focused around reducing management and process overhead in order to make the company more efficient, agile and responsive while reducing costs. The introduction of a process such as CMMI could easily double operational expenditures even without pursing official certification. It is also likely to reduce agility and increase ongoing development cost by reducing development team efficiency.
Lean Software Development is an application of the principles of Lean Manufacturing to the software development process. Lean is simply an attempt to eliminate waste from the development (or manufacturing) process and is directly related to concepts such as Just in Time Manufacturing – the process of having parts arrive at the factory just as they are needed for assembly reducing storage cost and leverage the time-value of money.
The concepts of Lean echo (or predate, in fact) concepts widely accepted in software development such as YAGNI (You Aren’t Gonna Need It) which encourage developers to save features that they cannot guarantee will be needed until they can. This often reduces the total amount of effort needed to make a product since useless codepaths are less likely to be constructed. Extra code not only costs at the time of creation but it also costs later to maintain and generally reducing system performance.
While Lean is often associated with process models and systems such as the CMMI and Six Sigma, Lean Software Development is more appropriately an Agile development methodology and is organized as such by its designers, Mary and Tom Poppendieck. Lean is a relative newcomer to the methodology space appearing as recently as 2003 and appears to be mostly a repackaging of most core Agile principles without the addition of many of the “scarier” practices such as pair programming in XP.
For any organization utilizing an agile process, Lean seems to be less of a firm methodology and more of a repackaging of obvious and common sense items. Lean’s greatest advantage, most likely, is for organizations who are caught in the “manufacturing” mentality to pick up a truly Agile framework and see it as “emulating Toyota”. Lean may have the ability to be sold to managers who otherwise won’t accept processes without management buzz words attached.
Six Sigma is a manufacturing process improvement program initially development by Motorola in 1986. Six Sigma and Motorola processes are exceptionally well known to businesses in Western New York as many of them, including then Motorola CEO George Fisher, were brought to Eastman Kodak. These manufacturing-designed processes and management when applied to Kodak, a research and development organization were used to drive out innovation, cripple agility, alienate the workforce and effectively dismantle the company.
Six Sigma itself is a manufacturing improvement process that focuses upon a feedback loop of quality improvement following on the tradition of many standard quality improvement procedures. In fact, Six Sigma has been criticized for being nothing more than marketing fluff applied to obvious principles long established in the industry. Six Sigma is often associated with other initiatives that have been labeled as management fads such as Total Quality Management and Reengineering. These processes and their ilk were lambasted by Scott Adams in his landmark business book The Dilbert Principle. Many of these fads and processes have come to be the face of poor management today even though mostly they have faded towards obscurity. Six Sigma remains more strongly and may not constitute a fad since it is simply a repackaging of traditional concepts.
The biggest problem with Six Sigma is that it is a manufacturing methodology that relies upon repeatable processes and measurable error rates to achieve its goals. As software development has neither of these things Six Sigma is meaningless in this space. Even Eastman Kodak, a company whose fortunes lay mostly in their research and development, saw Six Sigma and heavy manufacturing processes erode their competitive advantage when their organizational focus was shifted to low-profitability and easily off-shore-able plastics manufacturing rather than pharmaceutical and technology research. Even Motorola, the poster child of Six Sigma, saw its fortunes dwindle compared to its chip making competitors and the market as a whole once their focus shifted from new product development into manufacturing. In first world markets, any process that can be measured and improved through a system such as Six Sigma should be earmarked for third world consideration. In those markets, the principles behind much of Six Sigma should probably be applied to their manufacturing processes as there is much value when the core business is actually manufacturing and not product development and innovation.
The mistake often made with Six Sigma outside of manufacturing is the belief that all processes, even those that are inherently creative, can be meaningfully measured and quantified. Few people would think of applying the principles of Six Sigma to painting a picture, composing a symphony, writing a novel, acting in a play, dancing in the ballet, etc. Why? What about an individuals expression through dance is less quantifiable than the creation of software? Software, according to Donald Knuth, is literature and, according to Paul Graham, the processes making it are closer to art than engineering and even engineers would be appalled to find their engineering processes being measured and quantified as if they were factory workers – interchangeable cogs in a giant machine. If companies believe that design processes are interchangeable and replicable why are they not automating them. Clearly the end products of mechanical engineering and software development are both delivered as computer data files and so would be perfect candidates for total computer automation. Why has this not happened – Because both of these processes’ values are in their creativeness and not in their repeatability. Therefore Six Sigma has no meaning in this context.
Lean Six Sigma
In theory, Lean Six Sigma is the application of Lean’s Agile development principles to a Six Sigma process improvement environment. In manufacturing circles this is an obvious marriage. In software circles, however, Lean does not in any way explain how Six Sigma can be applied to the creative process of software design and writing. Even Mary and Tom Poppendieck, the creators of Lean Software Development, when speaking about Lean Six Sigma, are only able to explain the application of Lean in a software environment and neglect to explain the use of Six Sigma.
There are two things that often result from the application of process controls in an organization. The bad is the creation of unnecessarily heavyweight or even possibly harmful process that encumber or possibly disrupt productivity and increase overhead costs making it more difficult to achieve profitability. The good is the codification of common sense principles. What is difficult is to avoid the first while leveraging the second without letting the second become the first.
Few process management advantages come from applying unintuitive principles to software development. Sometimes principles which are obvious after you are aware of them, such as YAGNI, are non-obvious or even counter-intuitive to novice developers, managers and teams. In cases where the organization is not aware of these principles at an organizational level and needs to apply them down to teams it is generally effective to codify them. However, we do not believe that it is valuable to codify principles which are already understood and obvious.
Every organization has a different level of organization knowledge in software development. The larger the company the great need for process control and codification. Smaller companies have more opportunity to transmit good processes through the idea of Practice learning without the need for formal codification.
By transmitting through Practice we have an opportunity to teach, learn and grow rather than by enforcing. Transmitted knowledge gives everyone an opportunity to evaluate, internalize and comprehend procedures and, most importantly, it gives everyone the right to question them. If a process is codified within the organization then only a select few people are likely to have any real ability to question the applicability of a process or procedure and innovation may be stifled. Few processes are appropriate all of the time and so teaching everyone when they are useful and when they are cumbersome empowers everyone to be as effective as possible. Codification takes that decision out of the hands of the people with the most direct ability to determine the appropriateness of a procedure within a particular context.
Organizations blessed with highly skilled staff need very little process control as the staff will generally regulate themselves and follow meaningful processes on their own. Some process control is still necessary, in this there is no question. A minimally skilled organization, on the other hand, will require a great deal of process control to allow them to obtain any degree of productivity at all. This is the fundamental theory behind a great many business and management processes – the concept of “managing to the middle”. Managing to the middle means that instead of acquiring and training and empowering great staff the company has decided to acquire budget resources and using heavy process control to obtain some level of production from staff that would be mostly dysfunctional with those processes.
At CompanyX our base principles rely on the fact that we acquire great, motivated resources and empower them for innovation and creativity. This has been said many times in this text and it cannot be stated enough. Every management decision made at CompanyX must be done in the light of the fact that we trust our teams, and we believe in our teams. CompanyX is not a code manufactory. CompanyX creates literature. It creates art. It creates innovation. Therefore it is imperative that we refrain from utilizing process control beyond that which is absolutely necessary and, whenever possible, rely upon the foundations of Practice and professionalism to establish working processes that are themselves flexible and mutable.
In software development, risk is a constant presence that must be addressed. Risk comes in many forms such as choosing to enter an inappropriate marketplace, finding a new competitor after a project has begun, poor design, misidentification of business needs, bug-ridden software implementations, failed vendor support for key platform components, regulatory changes, unforeseen technical difficulties, team malfunctions, key employee health issues, lack of internal political support, etc. Risk is unavoidable and it is by facing risks that we are presented with an opportunity to move into new, unproven territories. With great risk comes great rewards. But nothing is without risk, even an absence of action does not truly mitigate risk, in fact it is likely to exacerbate it.
“If you don’t risk anything, you risk even more.” – Erica Jong
The first step in dealing with risk is identifying potential points of risk. For CompanyX we know that our most significant points of risk come from personnel dynamics which include health issues, availability, etc. and post-production market uptake. There are many other risk areas but to these two we are more susceptible than many companies and need to be acutely aware of the implications of these risks.
Before any project begins a detailed risk assessment process should be followed. This procedure should include development lead, design lead and financial management at a minimum. Our organization fortunately is able to avoid some areas of often high potential risk because of our minimal political turmoil, abundance of project support and lack of risk aversion. For example, CompanyX has never been forced to abandon a project because of a withdrawal of internal executive sponsorship which may plague even the most successful of projects in other companies.
Once a comprehensive set of foreseeable and reasonable risks are collected then we must assign a likelihood of occurrence and a level of impact quantity to them. Obviously this measurement is highly subjective.
Unlike a larger firm which may be able to spread risk out widely between projects and hedge high risk projects with low risk projects we are stuck in a position in which we must put our few eggs into very few proverbial baskets and our organizational risk to a single failed project is quite high.
Risk Mitigation and Handling
Once risks are understood there are four options for dealing with risks. Risk avoidance, reduction, retention and transference. Each approach must be considered and decided upon by the risk team. In general, however, it is in the interest of CompanyX to practice risk retention when possible within the software development practice because our ability to withstand failures internally is generally better than is the industry average causing other forms of risk mitigation to be more costly by comparison. My retaining risk we have a greater opportunity to leverage our internal skills more effectively. Obviously this does not apply when risk can simply be avoided.
CompanyX has pursued process management very diligently in the past and many of our processes and procedures reflect a dedication to research, customization and independence, but we still have a very long way to go. There is much that we have to learn from the industry and many best practices that we still need to apply.
Building a strong process supporting strong projects requires constant diligence, reevaluation and reassessment. There are always areas in which improvement is possible and must be pursued.
In this document we have identified areas of improvement that should be addressed such as our lack of organized testing and our lack of support for a testing Practice and a need to implement more formal risk assessment procedures. New software systems such as continuous integration and discussion forums needs to be installed and made available to the development teams.
In the past year we have been making many strides in the right direction and improvements are very visible. Perhaps our greatest achievement of the past year is the use of the wiki format for storing procedures, design documents and other forms of organization knowledge in a repository that is not only highly available but also easily searchable. It is very exciting to see how this resource will be adapted for use in the future.
Whatever process improvements we seek out in the future it is imperative that we remain focused upon the fact that our value lies in our staff and that any process which stymies their desire to create or hampers their ability to remain productive is dangerous and should be avoided. If we undermine our own value then we have lost the very resource for which we were hoping to be able to manage through process control. Like any form of management, process management has more ability to damage than it does to accentuate and care must always be taken.
Process management must be a tool for enabling innovation and creativity because it cannot be a tool to create these things.
Wells, Timothy D. Dynamic Software Development. New York: Auerbach, 2003.
Brooks, Frederick P. The Mythical Man Month: Essays on Software Engineering.
Graham, Paul. Hackers and Painters: Big Ideas from the Computer Age. Sebastopol: O’Reilly, 2004.
DeMarco, Tom & Lister, Timothy. Peopleware: Productive Projects and Teams. New York: Dorset House, 1987.
Rainwater, J. Hank. Herding Cats: A Primer for Programmers Who Lead Programmers. New York: APress, 2002.
Spolsky, Joel. Joel on Software. New York: APress.
Spolsky, Joel. More Joel on Software. New York: Apress, 2008.
Mason, Mike. Pragmatic Version Control: Using Subversion – The Pragmatic Starter Kit Volume 1. Raleigh: Pragmatic Bookshelf, 2006.
Hunt,Andrew & Thomas, Dave. Pragmatic Unit Testing in Java with JUnit – The Pragmatic Starter Kit Volume 2. Raleigh: Pragmatic Bookshelf, 2003.
Clark, Mike. Pragmatic Project Automation: How to Build, Deploy, and Monitor Java Applications – The Pragmatic Starter Kit Volume 3. Raleigh: Pragmatic Bookshelf, 2004.
Hunt, Andrew & Thomas, Dave. Pragmatic Programmer, The: From Journeyman to Master. New York: Addison-Wesley Professional, 1999.
Spolsky, Joel & Atwood, Jeff. Stack Overflow Podcast, Episode 23 – Seven Mistakes Made in the Development of Stack Overflow Segment.
Beck, Kent. Test-Driven Development by Example. New York: Addison-Wesley, 2003.
Beck, Kent & Andres, Cynthia. Extreme Programming Explained: Embrace Change. New York: Addison-Wesley, 2004.
Hass, Kathleen B. From Analyst to Leader: Elevating the Role of the Business Analyst – The Business Analyst Essential Library. Vienna: Management Concepts, 2008.
Cockburn, Alistair. Writing Effective Use Cases. New York: Addison-Wesley, 2000.
Ruby Module Test::Unit. http://ruby-doc.org/stdlib/libdoc/test/unit/rdoc/classes/Test/Unit.html
Vermeulen, Allan; Ambler, Scott W.; et. al. Elements of Java Style, The. New York: Cambridge University Press, 2000.
Baldwin, Kenneth; Gray, Andrew; Misfeldt, Trevor. Elements of C# Style, The. New York: Cambridge University Press, 2006.
Ambler, Scott W. Elements of UML 2.0 Style, The. New York: Cambridge University Press, 2005.
Knuth, Donald. “Literate Programming” in Literate Programming, CSLI. 1992.
Wikipedia Contributors. “Capability Maturity Model Integration”. http://en.wikipedia.org/wiki/Capability_Maturity_Model_Integration November, 2008.
Wikipedia Contributors. “Six Sigma”. http://en.wikipedia.org/wiki/Six_Sigma November, 2008.
Wikipedia Contributors. “Lean Software Development”. http://en.wikipedia.org/wiki/Lean_software_development November, 2008.
Wikipedia Contributors. “Risk Management”. http://en.wikipedia.org/wiki/Risk_management November, 2008.
Adams, Scott. Dilbert Principle, The. New York: Collins, 1997.
Poppendieck, Mary; Poppendieck, Tom. “Why the Lean in Lean Six Sigma”. http://www.poppendieck.com/lean-six-sigma.htm November, 2008.