Monday, July 28, 2014

The Three Laws of Software Development - Conway's Law

During a recent class, Peter Saddington referenced three laws which collectively are knows as the Three Laws of Software Development:
  1. Humphrey's Law
  2. Ziv's Law
  3. Conway's Law
I wrote about Humphrey's and Ziv's Laws in my previous posts. Conway's Law is perhaps the one that is most well known of the three.
Looking up "Conway's Law" you'll find it in Jeff Sutherland's book "Exploring Scrum: The Fundamentals." Conway's Law is an adage named after Melvin Conway who first mentioned the idea in 1968. It states "Organizations which design systems...are constrained to produce designs which are copies of the communication structures of these organizations" or it's also written as "If you have four groups working on a compiler, you'll get a 4-pass compiler" (sic). That second version is a bit dated by today's standards. I believe Peter stated it a bit more colloquially as "Software always mirrors the group that built and designed it - if the company is dysfunctional due to poor communication then the software will have the same dysfunction" or simply "everyone needs to communicate better."
This law is as proverbial as the previous two and like them, is more of a valid sociological observation. I found this diagram by Manu Cornet on Bonkers World that I thought was pretty funny and actually a fair representation of Conway's Law...

Manu's diagram isn't specific to Conway but you can infer what you like. The Microsoft and Oracle diagrams are especially telling. Peter tied this law into the need for excellent communication and collaboration as a necessity in producing great software. I've personally seen this law represented myself in the many start-up companies that I've had the honor of working with and the opposite end of the spectrum, large corporate software endeavors (at some point I'll relate a project that had over 100 sign-offs).
Early stage Start-ups are typically small and have a tiny organizational chart - the software produced is usually fairly simple structurally and the communication is more one-to-one due to the tight org structure. This ability to be nimble due to size produces many advantages, especially the ability to go to market quickly. Larger organizations begin struggling to produce software with any speed - there are so many check-points and stakeholders involved in the production that there are often small changes that impact both the software in complexity and the company's ability to market quickly. Large corporate entities are often bound to many revisions, checks-and-balances, approvals - you get the idea. Not only does the software fill with complexity but releases become a nightmare.
Tying back to the class, the emphasis is to make small incremental changes and expose those changes for approval back to the stakeholders early and often so the team can adjust and do a better job meeting expectations. Not only do you produce better and more efficient software, but the release becomes less problematic - there are fewer surprises. I don't know how many times I've worked on a project where the smaller changes truly delight the customer while the larger changes produce confusion and revolt - this is usually ascribed to poor training but even taken that as a viable excuse, doesn't it amount to a lack of communication?

No comments:

Post a Comment