Why Traditional Software Design Fails in the Age of Complex Business Platforms

Introduction

Alibaba Vice President Xuan Nan explains that in today’s rapidly changing business environment, the only constant is change. To keep infrastructure secure, highly available, and supportive of fast innovation, software architecture must evolve.

Note: The term “software” here refers to end‑user business applications, not operating systems or middleware.

1. Stages of Software Development

Tools & Information Era – early numerical computing tools such as Excel, Word, and internal management systems.

Internet Service Era – the rise of web services like Taobao, QQ, Google, and Sina.

Social Infrastructure Era – mobile Internet, IoT, and AI make software a fundamental part of society, forming a complex ecosystem.

These shifts challenge traditional system architecture and organizational models, moving from quantitative to qualitative change.

Alibaba as a Case Study

From a single Taobao platform to dozens of business units (new retail, cloud, entertainment, health, etc.) that evolve rapidly.

From a monolithic Denali system to a cluster of tens of thousands of distributed services.

From a team of dozens to tens of thousands, each working with limited visibility of the whole.

From stoppable releases to continuous, non‑interruptible services that must guarantee both system stability and business continuity.

2. Traditional Software Design – External Determinism

Programmers seek certainty; changing requirements are seen as a fatal flaw. In the “tool & information” era, clear business boundaries allowed waterfall development with heavy documentation.

With the rise of internet services, agile methods accelerated iteration but often resulted in patch‑heavy codebases, rapid architectural decay, and loss of reliable documentation.

Two main sources of change are identified:

User‑scale growth – addressed by distributed databases, caches, services, multi‑datacenter deployments, CDN, etc.

Business‑function evolution – handled via large‑field databases, NoSQL, metadata + key‑value stores, workflow engines, rule engines, UI componentization, and plugin architectures.

Traditional design assumes a deterministic software boundary, using abstraction, modularity, and configurability to cope with change.

3. Changing Perspective – Internal Determinism

As software becomes societal infrastructure, its boundaries blur. The article likens this shift to moving from traditional machine‑learning models (which require a well‑defined problem space) to deep‑learning models that define a stable core structure and learn from data.

Software engineering must therefore find a stable “core” (the kernel) and let the outer layers evolve, using deductive reasoning for the core and inductive reasoning for extensions.

4. Documentation as Code

Historically, each programming language abstraction improved productivity, but the source code remains the most valuable artifact. Modern business platforms often lack documentation that reflects core business concepts (e.g., transaction models, marketplace rules).

The author argues for merging design and implementation so that documentation lives inside the system, similar to DNA in living organisms, ensuring it cannot disappear.

5. From Component‑Based to Object‑Based Design

Inorganic systems are built top‑down from standardized components; organic systems grow from the inside out, guided by “genes.” Applying this metaphor, business platforms should adopt object‑oriented designs where each business unit (e.g., Taobao, Tmall) is represented as a class inheriting from a common parent, enabling independent evolution while preserving core invariants.

Collaboration mechanisms between business objects (e.g., Taobao ↔ Tmall, Damai ↔ Fliggy) create a dynamic ecosystem.

The evolution of Alibaba’s business platform moves from function‑centric to capability‑centric, and finally to business‑centric architecture.