Random Acts of Architecture

Few things are so error prone and political as the architects’  “technical vision”. Technical vision identifies longer term technical goals (such as add iPad support or upgrade to the new database version) and describes how to reach them, perhaps over multiple releases. Clearly, this is only useful for larger, ongoing or technically complex projects.

Many architects aim for isolated technical improvements in their first technical vision. Unshackled from the constraints of time and resources, these are often ambitious, large-scale changes intended to make anticipated subsequent changes easier by adopting new technologies, creating reusable components and standardizing formats and protocols.

The first issue with this approach is that little is tied back to customer or business requirements. Otherwise, the architect risks an “ivory tower” situation, recommending or enforcing designs often blamed for inflated resource requirements and missed deadlines.

The best way to avoid this and gain credibility is to interview stakeholders, identify changes they want made and incorporate them into the vision. Non-business stakeholders like testers, localizers and documentation writers often want product improvements, too. Perceptive architects prioritize stakeholder requests that echo their own thoughts and, once the architect has demonstrated delivering these in products, architects will gain the credibility and latitude to add some of their own ideas.

The second issue is with component reuse and standardization. While the architect’s high level view is good for identifying opportunities for these and prioritizing them, links back to business requirements are difficult to articulate, particularly to non-technical stakeholders. Developers hold code written by others to a higher standard than their own (particularly documentation, quality, scalability and security), will use such components in ways never originally envisaged or intended and backwards compatibility requirements can limit improvements. However, the biggest danger with writing reusable components and standardization is the tendency to focus on them and not the problems they solve.

Writing reusable components and standardization are organizational decisions as much as architectural ones. Bugs or enhancement requests need tracking, prioritizing, fixing in a timely manner and distribution. Good approaches for reusable components include starting with small, non-critical components or creating internal “open source” projects where anyone can contribute, subject to review. Avoid forking components if possible.

Standardization requires a middle ground between forgoing features (“lowest common denominator” approach) and spending additional effort (the “kitchen ink” approach) and organizations must resist doing too much of the latter. It is best to start with often used interfaces and include security, scalability, performance, error behaviour and similar non-functional requirements.

Lastly, new technologies are often attractive, exciting and claim to solve problems more effectively, reducing development time and cost. However, adopting new technology requires an understanding of the whole system and not just isolated problems, lest it cause more problems than it solves. Once again, there is an organizational cost with retraining developers, purchasing new tools and integrating it with old technologies. New technologies are best trialled first with a proof of concept.

Good technical visions are known to and have the backing of both stakeholders and developers, which is easier if creating it involved them. Technical visions need not be detailed or specific, at least not initially and good technical visions are accessible, unambiguous and concise. If others cannot read them or understand them, it cannot be implemented or implementations will digress from the intention. As mentioned before in this blog, it is the architects’ responsibility to communicate the vision effectively.

Technical visions can change over time, too. This reflects changing business requirements, technology landscape and development techniques. It also reflects expressing the intention more effectively. Constructive criticism and discussion is vital, both to improve it and, as also mentioned before in this blog, incorporating others’ feedback gives them ownership and ensures it is current.