Software development technologies: the stack explained, and how to choose
Software development technologies are the languages, frameworks, databases, cloud services and tools a system is built and run with. Choosing them is one of the first decisions in any software project, and one of the longest-lasting: the stack decides who can maintain the system, how long it stays supported and what it costs to run. This guide explains the layers of a modern stack, which technologies are most used, how long each is supported, and how to choose without following fashion.
It is written for business owners, managers and product people who need to understand or approve a technology choice, not to write the code themselves. For how projects are organised, see our guide to software development methodologies.
The short version
- A business system combines several technologies, one or more per layer: interface, backend, data, infrastructure and delivery.
- Established technologies with long support, large communities and many people who know them are usually the right choice for core systems.
- Every technology has a support lifecycle; plan upgrades before a version runs out, not after.
- Check the licence of every component; it decides what you may do with the software you paid for.
- The best stack is the one your future maintainers can work with, not the one that looks newest.
The layers of a software system
Most business software has the same layers, whatever it does. Each needs its own technology choice:
| Layer | What it does | Common technologies |
|---|---|---|
| Web interface | What users see and use in the browser | HTML, CSS, JavaScript and TypeScript; frameworks such as React, Vue, Angular and Svelte |
| Mobile apps | Apps installed on iPhone and Android | Swift for iOS, Kotlin for Android, or cross-platform Flutter and React Native |
| Backend | Business rules, APIs, sign-in, jobs | Node.js with TypeScript, Python, Java, C# with .NET, Go, PHP |
| Data | Storing and querying the data | Relational databases such as PostgreSQL, MySQL and SQL Server; SQLite on devices; Redis for caching; document stores such as MongoDB |
| Infrastructure | Where and how the system runs | Cloud platforms such as AWS, Azure and Google Cloud; containers with Docker; Kubernetes; infrastructure as code with Terraform |
| Integration | Connecting to other systems | REST and GraphQL APIs, webhooks, message queues, file exchange |
| Delivery and quality | Building, testing and releasing changes | Git, continuous integration and deployment pipelines, automated tests, monitoring |
| AI features | Language models, search, prediction | Model APIs from AI providers, open models, vector search such as pgvector in PostgreSQL |
Not every system needs every layer. An internal tool may have only a web interface, a backend and a database; a consumer app may lean on a backend-as-a-service. Our backend app development page explains that choice for apps, and Node.js app development covers one common backend stack in detail. For how the layers map to kinds of projects, see types of software development.
Which technologies are most used
The largest public measure of technology use is Stack Overflow's yearly survey. In the Stack Overflow survey 2026, Technology (read 11 November 2026), JavaScript was the most used programming language for the twelfth year in a row. Among those who answered the language question:
| Language | Share of respondents |
|---|---|
| JavaScript | 62.0% |
| SQL | 58.4% |
| HTML/CSS | 58.2% |
| Python | 58.0% |
| Bash/Shell | 51.3% |
| TypeScript | 43.8% |
| Java | 27.6% |
| C# | 27.1% |
| PowerShell | 24.6% |
| C++ | 24.3% |
For databases, PostgreSQL led for another year, used by 57.9% of those who answered that question, ahead of SQLite at 42.1%, MySQL at 35.1% and Redis at 28.3%. Two cautions apply. The survey counts people who chose to answer, mostly from Stack Overflow's own audience, so it measures popularity among practitioners rather than the market. And use is not suitability: a language can be widely used and still be the wrong choice for your system. But popularity has practical value for a buyer: more people who can maintain the system, more libraries, faster answers to problems, and less risk that the technology fades.
Support lifecycles: the part buyers forget
Every language, runtime, framework and database version stops receiving security fixes at some point. Software built on an unsupported version keeps working, until a vulnerability is found that nobody will fix. The policies differ, so they belong in the plan from the start (all read 11 November 2026):
| Technology | Support policy | Example |
|---|---|---|
| Python | A new version every October, each supported for about five years | Python 3.11 reaches end of life in October 2027, 3.12 in October 2028, 3.15 (released 9 October 2026) in October 2031 |
| Node.js | Long-term support releases typically get critical fixes for 30 months; production should use only LTS releases | From Node.js 27, the cycle becomes annual and every major version moves to LTS |
| .NET | A release every November; even-numbered LTS releases get three years, odd-numbered STS releases two | .NET 10 (LTS) and .NET 11 (STS, released 10 November 2026) are both supported until 14 November 2028; .NET 9 ended on 10 November 2026 |
| Java | Oracle provides long-term support on specific releases, planned every two years | Java 8, 11, 17, 21 and 25 are LTS; the next planned LTS is Java 29 in September 2027 (roadmap updated 20 October 2026) |
| PostgreSQL | A new major version about once a year, each supported for five years | A major upgrade every few years keeps a database inside support |
Frameworks and mobile platforms add their own schedules: Apple and Google raise their minimum requirements for app updates every year, and web frameworks retire old major versions. A maintenance plan lists every component with its end of support date and schedules the upgrades, which is part of what our mobile app maintenance covers.
Licences decide what you own
Almost every system is built mostly from open-source components, and each comes with a licence. Permissive licences such as MIT and Apache 2.0 let you use, change and sell software that includes the component, with few conditions. Copyleft licences such as the GPL require that software distributed with the component is shared under the same terms, and the AGPL extends that to software offered to users over a network, which matters for SaaS products. Some components are offered under a commercial licence and an open-source one, and the commercial terms apply to most business use. The list of components and licences should be part of what you receive at handover.
The EU now requires that list for many products. The Cyber Resilience Act, Regulation (EU) 2024/2847 (read 11 November 2026), requires manufacturers of products with digital elements to identify and document the components they contain, including by drawing up a software bill of materials in a commonly used, machine-readable format covering at least the top-level dependencies (Annex I, Part II, point 1). Its main obligations apply from 11 December 2027. Choosing well-maintained components with clear licences makes that requirement simple.
How to choose software development technologies
A good technology choice answers practical questions, not "what is newest":
- Who will maintain it? Your own team's skills, or the skills you can hire in your market, count for more than any feature. A system only one supplier can work on is a risk.
- What does the domain need? Some areas have a clear leader: Python for data and machine learning, Swift and Kotlin for apps that use the newest platform features, relational databases for business data with relationships.
- How long must it run? A system meant to run for ten years needs technologies with long, predictable support and a history of careful upgrades.
- What must it connect to? Existing systems, identity providers and the platforms your customers use often favour one stack over another.
- Where must it run? Hosting region, data location and any customer requirements for specific clouds or on-premises installation.
- What does it cost to run? Licences, cloud services and the people needed to operate it, over years rather than at launch.
Fewer technologies are usually better than more. Each extra language, database or platform adds upgrades, security patches and knowledge that someone has to keep. A well-known stack, used consistently, is easier to staff, cheaper to maintain and easier to hand over.
Common mistakes
- Choosing by trend. A technology that is new this year may be abandoned in three; the system will outlive the trend.
- A different stack for every project. Each one needs its own skills and upgrades; the company ends up maintaining a museum.
- Ignoring end of support. The upgrade becomes an emergency project when a vulnerability appears in a version nobody patches.
- Locking into one vendor without deciding to. Proprietary platforms can be the right choice, but the cost of leaving should be known before choosing.
- No list of components. Without it, nobody can say whether a newly announced vulnerability affects the system.
What we use, and why
At DT Nova we keep a deliberately small set: TypeScript for web interfaces and many backends, Swift and Kotlin or a cross-platform framework for apps, Python where data and AI work need it, .NET where a client's systems already run on it, PostgreSQL for data, and containers in an EU region of a major cloud. Each project's scope names the technologies with the reasons, the support dates and the licences, so the choice can be checked by anyone, and the code is in the client's own repository from the first commit. For the services built on them, see custom software development and mobile app development; for a second opinion on an existing stack, software consulting.
Questions people ask
What are software development technologies?
The languages, frameworks, databases, cloud services and tools a software system is built and run with. A typical business system combines several: a web or mobile frontend, a backend language and framework, a database, a cloud platform with containers, and tools for version control, testing and deployment. Together they are called the technology stack.
What is the most used programming language?
In the Stack Overflow survey 2026, JavaScript was the most used language for the twelfth year in a row, named by 62.0% of those who answered the language question, followed by SQL at 58.4%, HTML/CSS at 58.2% and Python at 58.0%. Popularity matters for hiring and libraries, but it is only one of the criteria for choosing a stack.
How long are programming languages and frameworks supported?
Each has its own policy. Python versions get about five years from first release; Node.js long-term support releases typically get critical fixes for 30 months; .NET long-term support releases get three years and standard releases two; Oracle offers long-term support on Java 8, 11, 17, 21 and 25; and PostgreSQL supports each major version for five years. A system should be planned to move to a supported version before the old one runs out.
Should a company choose the newest technology?
Usually not for its core systems. Mature technologies with large communities, long support and many people who know them are cheaper to maintain and easier to hand over. New tools earn their place where they solve a real problem the established ones do not, and they are best tried in a contained part of a system first.
Does the technology choice affect who owns the software?
It can. Open-source technologies under permissive licences such as MIT or Apache 2.0 leave you free to use and change the code. Copyleft licences such as the GPL or AGPL carry obligations when software is distributed or, for the AGPL, offered over a network. Proprietary platforms can tie the system to one vendor. The licences of every component should be checked and listed before a system is built.
Sources
- Stack Overflow survey 2026, Technology (read 11 November 2026)
- Python devguide, Status of Python versions (read 11 November 2026)
- Node.js, Node.js releases (read 11 November 2026)
- Microsoft, .NET and .NET Core support policy (read 11 November 2026)
- Oracle, Java SE Support Roadmap (updated 20 October 2026, read 11 November 2026)
- PostgreSQL, Versioning policy (read 11 November 2026)
- Regulation (EU) 2024/2847, Cyber Resilience Act, Annex I and Article 71 (read 11 November 2026)