There are no universal notations, methods or techniques for software engineering. This is because different types of software require different approaches.
There are cases where software projects or software can fail.
As new software technology helps us to build bigger, more complicated systems, the requirements change.
Systems need to be developed and delivered faster meaning more complicated systems are required. This means more requirements for the systems.
It is easy to develop programs without using software engineering techniques… BUT This can result in more expensive software development and less reliable, readable systems. To solve this issue, more education and training is required on these techniques.
This is where the clients and the developers define the software that should be produced. This is where the software is limited in terms of what is should be.
This is where the software is designed and programmed.
This step ensures that the software does what the client requested.
When software is modified to meet client’s or the market’s new requirements.
A software process model is a set or related activities that leads to a software product. An abstract descriptions of high level software processes. These can be used to explain different approaches to software development. These can be interpreted as frameworks that can be expanded and personalized to create more specific software engineering processes.
The model takes the fundamental software process activities. Seperate process phases: - Requirement specification - Software design - Implementation - Test
This approach combines the activities with specification, development and validation. The system is developed in a series of versions (increments) Each version adds some functionality on top of the previous version.
This approach depends on the availability of reusable components or systems. The systemdevelopment process focuses on configuring these components to be used in new contexts and integrate them with the system. This lets you reuse code from previous projects and save time and money.
Activities are done in a sequence. Handover of workproducts between phases and milestones are used to monitor progress. Waterfall model is an example of a plan-driven process. You are supposed to have all of the process activities and phases planned out before you start development. The phases in the model directly reflect the fundamental development activities.
The systems services, limits and goal and made concrete through consulting with the users. They are then defined in detail and are used as the system specification.
The overall system architecture is established in this phase. Identify and describe the fundamental system abstractions and their relations (UML)
The design is implemented in the indivivual program parts. The design is tested to verify that it meets the specification.
The individual program parts are integrated and the whole system it tested. The software is delivered to the client.
The system is now installed and in use. The system is maintained to any issuse that weren’t found earlier. The system is improved over time.
For software that needs to be flexible while it is being developed.
The software must interface with hardware. In this case the hardware is not flexible, thus the software must be.
When total security is a requirement of the specification and design. These systems must have finite specifications and design documents.
Part of a larger system being developed by several parties. This makes finite specifications extremely necessary.
When one phase ends another begins. Steps come ordered and don’t allow for going back and redoing parts. (waterfall flows only down) When the requirements are mostly unchanging.
Activities in sequence versus all activities at the same time. Agile development is very flexible and the requirements may be constantly changing.
You can iterate within increments and the increments can be planned. You work within fixed time slots and update the project backlog continuously. Incremental is an agile process, where you end with multiple versions that you can show the clients as the project progresses. This allows for the client to come with feedback along the way.
Development risk is reduced by reuse but there is the risk of not being able to make the desired changes at all or in the time frame. Most projects have some level of code reuse. This is often informal. This reuse requires looking for the existing code, changing them to meet the requirements and integrating them with the new code.
The incremental development is based on the idea that: 1. Develop in prototypes. 2. Get feedback from the users and others. 3. Develop over multiple versions until the required system is produced.
Incremental development is the most common approach to developing applications and software.
Yes, it can be either one or a mix of both.
Identify the system increments in advance. A predictable waterfall plan is split into parts.
The early increments are identified. The later increments depend on progress and the clients priorities. You work in fixed timeframes and update the full project backlog as you go.
Focus on the individuals and teamwork rather than the processes and tools. Good software comes before comprehensive documentation. Work with the client instead of using contract work. Deal with changes instead of sticking to the plan.
Development costs are reduced The amount of analysis and documentation that has to be redone is much less than waterfall.
It is easier to get client feedback The clients can comment on demonstrations of the different versions and see the progress.
Earlier delivery of new software New features can be made available even if they are not fully completed.
The project is not visible Managers can have difficulty measuring progress. It would be a waste of resources to produce documents that reflect the different versions.
The system structure can get messy over time Changes lead to messy code. It gets increasingly difficult to add new functions to a system. Large, complex systems with large teams struggle with incremental for this reason. Large systems need a stable architecture. Responsibility of the different teams needs to be clear with respect to this architecture. This has to be done in advance.
The inital requirements are suggested They don’t need to be developed in more detail. They should include short descriptions of important requirements and desired functionality.
With the overview of the requirements, components are searched for that provide the necessary functionality.
The requirements are polished with the knowledge of the reusable components that were found.
If an application that meets the requirements is available, it is configured for use in the new system.
This model reduces the amount of software that must be developed. This in turn, reduces the costs and risks.
You usually don’t have control of the software that is being reused. This can include for example how and when new version are released and how the functionality is changed.
Plandriven means the desired result can be predicted. Plandriven = waterfall. Plandriven is an approach where the development process is planned in detail. A project plan is created that registers the work that needs to done, who should do it the development plan and the work tools. Managers use this plan to support project decisions and as a way to measure progress. This is a traditional approach to software development.
Agile expects changes and frequent user inspection to get the best results. Agile = Iterative, more detailed Scrum and XP Agile methods are iterative, the software is developed and delivered in stages. These versions are not planned in advance but are chosen underway. Decisions on what should be included in a version depend on the clients priorities.
Goal: to handle changes in the project. Small teams. Environment is turbulent and fast-paced, project focused.
Goal: predictability, stability and security. Larger teams and projects. Environment: stable, few changes, organisation focused.
Customer relations: dedicated clients on site, focused on prioritised changes Planning and control: qualitative control. Who and how doensn’t matter as long as it gets done. Communications: Tacit knowledge. People do things without needing much explanation or discussion.
Customer relations: More formal and infrequent. Focused on contract decisions. Planning and control: Documented plans, quantitative control. It is important to know who does what. Communications: Explicit. Plans must be discussed and verbalized and shared with the others.
Requirements: can withstand unpredictability. Development: simple design, small increments, refactoring is assumed to be cheap. Test: Test cases define the requirements.
Requirements: Formal project, user interface, quality, predictable requirements. Development: Comprehensive design, larger intervals, refactoring is costly. Test: Documented testplans and procedures.
Cockburn characteristics can describe a programmers personality. One type may be more favourable than another depending on agile/plan-driven approach.
As soon as the work on a task is complete, it is integrated into the whole system. After such integration, all the unit tests must pass. Continuous integration uses tools to automate the process. Depends on unit tests. Does NOT remove the need for tests.
Openness of all work. Respect each other. Focus on the common goal. Courage for difficult decisions. Duty to the common goal.
3 roles. 5 events. 3 artifacts.
Transparency: Everyone knows what needs to be done and who is doing what. Inspect: Keep an eye on where we are heading (daily meetings). Adapt: Change if it is necessary.
Commitment: to reach the sprint goal. Focus: on what needs to be done in the sprint. Openness: Communication is key, don’t hide issuse. Respect: for each other Courage: to do the right thing
Scrum master is a manager. No communication with the client. New tasks are added to the sprint backlog in the middle of a sprint.
Inspect progress towards the sprint goal. Adjust the backlog accordingly. Update (how far are we?) Short meeting (15 min) Tell the others if you need help.
Work together with the whole SCRUM team for sprints. Look in the backlog. Make a sprint backlog. (this can’t be changed from the outside)
Check development status. Sprint review max 4 hours with sprints of 4 weeks. The client participates along with the team. Only talk about what has been done. Dialogue, no presentation. Update the backlog.
For increasing quality and effictivity. What went well/bad? How can this be improved for the future?
Stakeholder contact Updates/prioritises the product backlog Can be a person dedicated to the client The goal is the maximise product value Can delegate but is responsible
Ensure everyone is keeping in-line with SCRUM Not the management leader Ensure the team is effective Deal with the teams blockages
They own the backlog Programmers, UI, UX and so on
An agile, incremental development method with focus on: - Collaboration - Quick and early software creation - Skillfull development practices XP takes all the ‘good things’ the extreme. (testing, pair programming).
The customre should be available full time for the use of the XP team. In XP, the customer is a member of the dev team and is responsible for bringing requirements.
Pair programming: developers work in pairs, checking each other’s work to ensure quality.
Pair-programming. Writing unittests before the code (with the help of TDD). Partners often have to integrate their code (use continuous integration). Refactor as often as possible. Collective ownership of code. Customer on-site, user stories, planning game, …
Work should be done incrementally/iteratively. Teamwork, transparency, communication and prioritisation are crucial. Requirements are broken down into bite-size pieces. There is overlap with the roles. (XP client and SCRUM’s product owner)
Communication, simplicity, feedback, courage.
Pair programming Customer on-site Acceptance test Daily standup (short meetings)
Teams implement exactly what was asked for. Nothing more. Strive for simple designs and quality code.
Early and frequent feedback is crucial. Feedback can come from unittests, team members and the client. Continuous integration. Acceptance test that the client performs. Short sprints.
Developers should be honest. Don’t make excuses for issuse. Don’t be afraid to make big changes.
E.g. If tests are good, do them all the time. Takes all good things, and places them at the core of the process.
Unit test, pair review, customer on-site, continuous integration, testing, early test, unit test, TDD.
Brief feature request, a promise for conversation. Written on a card with criteria for confirmation on the back.
To see the bigger picture of the user stories. To understand how things are now and to imagine how they could be. Visualize the stories you tell about your software.
Story maps consist of: * User - a card that tells a story about a type of person, doing something to reach a goal. * Activities - Jobs done by similar people to reach a time * Backbone - Activies and jobs on a higher goal tier, give the user story structure. - This is a big goal, that the little goals are attached to. * User tasks - Short, concise sentences that explain the goal. * Sub-tasks - Break down more complicated goals. * Release slices - Identify tasks. The smallest number of tasks that allow specific users to reach their goal.
Story map process has 4 levels
“As a
Exploration, planning, iterations to first release, productionizing, maintenance.
Iteration
Elicitation and analysis of needs. Specification of requirements. Validation of requirements.
There are two fundamental approaches to Requirement Elicitation: 1. Interview, where people talk about what they are doing. 2. Observation or etnography, where you observe how people do their work and which technologies they use and so on.
Use a mix of interview and observation to gather information. This can be used to find the requirements that form the basis of further discussion.
Is a process in which you write down user and system requirements into a document. Ideally, these requirements should be clear, consistent, complete and easy to understand. User requirements should be written in natural language and supplemented with dialogue and tables in the document. System requirements can also be written in natural language, but other notations, graphs, maths, etc can also be used. (state machines, automata)
Is the process of controlling of ensuring that the system requirements will are really what the client wants. There are different checks that can be used to validate the requirements.
Validity checks: Check if the requirements reflect the users real needs. User needs can change over time, so this is an important thing to keep track of.
Consistency checks: Requirements should not conflict with others.
Completeness checks: The requirement specification should be comprehensive for every function and the limits that the client wants.
Realism checks: Using knowledge on existing systems, control the requirements to ensure that they fit within the budget and time-frame.
Verifiability: You should be able to create tests that verify whether or not a requirement is met.
Requirement reviews Requirements are analyzed systematically by a team of judges, to check for mistakes or conflicts.
Prototyping Develop executable models of the system and verify with the client that it meets their expectations.
Test-case generation Tests can be designed along with requirements instead of after the fact. If it is difficult to design tests for a requirement, that can mean the requirement is unrealistic.
Discovery & Classification This is the process of interacting with the stakeholders to find their requirements.
Categorization Take the unstructured list of requirements and group related requirements together.
Prioritization & Negotiation Conflicts can arise when there are multiple stakeholders. Prioritize the most important requirements, through negotiation, discussions, compromises and meetings.
Documentation Here the requirements are documented.
Many stakeholders with conflicting needs. Stakeholders speak their own “language”. Lack of communication because things are assumed to be “obvious”.
Stories, Scenarios.
Waterfall: Verify the requirement specificatino with strict change management. SCRUM: Product vision and product backlog, are discussed and updated every sprint. Product Planning: Product vision, release plans and/or product roadmaps. XP: User Stories.
Waterfall: Be up front in the requirements phase - state it now or it will be difficult later on. SCRUM: Ongoing refinement of product backlog with stakeholders, say what is most important now, we will continue. XP: Customer on-site.
User personas are a useful technique to describe users of your product. A fictional character with a name, picture, relevant characteristics, behavior, opinions and a goal. Different people can have different goals. Understand a personas goal is useful for creating a product that is meaningful to the users.
Short description on a type of user, a goal and a reason.
Product Feature: Corresponds to an Epic. Epic: A collection of related user stories. User story: Breakdown of an Epic.
The experiences a person has, when interacting with the software.
This is to do list of items a SCRUM team must tackle. - Software requirements. - User stories. - Descriptions of supplementary tasks that are needed, such as architecture definition or user documentation.
A risk is a potential problem. The possibility of loss or damage. Risk Management: project leaders must evaluate the risks that can affect a project, monitor them, and handle them when problems arrise.
Risks that threaten the project plan. Time will be wasted and costs will rise.
Architectural design. Arrises because problems can be harder to solve than expected. Vagueness in the specification. Project gets older and starts to decay.
Market risk. What if no one uses the product? Strategic risk. We don’t need that new component after all. Sales risk. How the fuck do we sell this?! Management risk. The top management don’t support the project anymore. Budget risks. Budget or personnel is lost.
Risk analysis and management are actions, that help a software team understand and handle uncertainty. It is a good idea to identify risks. Evaluate the probability of risks. Estimate the impact of a risk and form a reaction plan for if the risk actually happens.
Risk exposure = probability * loss, describe consequence. Probability < 100%. If p = 100% then it’s an issue.
Prioritize according to risk exposure, establish cut-line. Deal with the risks above the line, accept the ones below.
Establish for each risk above the cut-line (RMMM: Risk Mitigation, monitor, management)
We want to prevent the risk from becoming an issue. We can reduce the probability. Or try to reduce the associated loss. Risk exposure = probability * loss.
For when a risk has become a loss, try to minimize the loss. This assumes the mitigation activity was unsuccessful. This is done by the project leader.
Observe how risks change over time. How the probabilities, loss, or the environment, change over time.
Risk and risk plans are part of the plans in project management. Development of others plans to contribute to identification of risk. It is planned.
Daily SCRUM: Do you have any impediments? Sprint review: Inspects risk related to product and stakeholder. Sprint retrospective: Adresses risks related to how the team works.
Personal shortcomings, unrealistic schedule, wrong function…
Safety, security, reliability, complexity, adaptability, testability, understandability, efficiency, usability, etc…
Quality is evaluated aesthetically, symbolically and functionally Quality can be either objective or subjective. Quality may not always be obvious.
Quality is a reflection of one or more peoples evaluation of the compliance of a product or service with their expectations. Quality can be broken into three types of categories: 1. Product quality. 2. Process quality. 3. Quality of expectations.
Quality tradeoffs are unavoidable. Quality consists of: - Quality assurance: plan or design processes to prevent bad quality. - Quality control: track that work products meet quality standards.
Cost of not doing it is bad quality - fixing errors.
Direct cost of error correction: - Loss. (effort) - Wasted work. (for users of the program) - Maintenance usually has larger costs than development.
Indirect cost of error correction - Follows from poor quality (unsatisfied users) - Has potentially severe consequences (losing customers)
Quality Management reduces these costs significantly.
Are we building a system that is fit for use? Compliance with the users expectations and experiences?
Do we pass all tests and requirements? Are we building a system with all the requirements implemented? Unit/integration tests
Testing: of programmes and prototypes. Review: of specifications, documentation and programs.
A user must participate in order to validate. Verification focuses on the compliance to the specifications and a client usually doensn’t participate.
Tests are a set of practices that support verification and validation. The purpose is to ensure a program does what it is supposed to, and to discover errors before delivery. This is done by making sure the progrm meets the requirements and by finding incorrect or undesirable behaviour. Verification: Unit test, component test. Validation: prototype test, user acceptance test.
Evaluation of work of one or more people with similar skills (peers). Mostly in the form of documents but can also be analysis of code.
Review is static, and there is no interaction between errors found in review. Tests are dynamic and errors can come as side-affects of an initial errors. Reviews (inspections) and tests are complementary to quality techniques. Both should be used under the verification and validation process. Inspections can control compliance with a specifications but not with the clients or users actual requirements. (Unless the user participates in the review. Prototypes are preferred for user participation) Inspecions cannot control non-functional properties such as performance, usability, etc.
For documents, designs, architectures, plans, etc.
For functionality and dynamic use of the program.
A model that shows the connection between tests at different levels and primary activities that drive the tests.
Unit test, component test, integration test, system test, user acceptance test.
Unit test: confirm valid and invalid input. Integration test: confirm that interfaces are compatible and work as expected. Acceptance test: validate fit for use, exploratory test.
Plan driven: in the end (often a dedicated test-team aspart of QA) Agile: all the time (test competence on the team, accept criteria on story, automated test, TDD)
DevOps is a method for both development and operation. DevOps is a development method for IT systems that connects different activities in projects. DevOps is a culture, that focuses on the entire software productions life cycle. The goal is to remove barriers between development and operation teams, to be able to react quickly to the users needs. It is also defined by The Three Ways: 1. Flow. 2. Feedback. 3. Continuous Learning.
When the code is checked, it is automatically integrated with the system. Speed up the rate of delivery and run tests constantly. Bsed on tools to automate the process. Depends on a suite of unit tests. Does NOT eliminate the need for testers.
Continuous Testing in DevOps is a type of software test that involves testing at all stages of a develoments lifecycle. The goal is the continuously evaluate the quality of the software.
Ensure that code can be implemented securely. Ensure that the business and service application function as expected and deliver every change to production.
Ensure that tests are automated and that every change is automatically implemented in production. Makes the development and release process faster and more robust.
Automated access to well defined environments. Tools like Docker for containerization or Virtual Machines.