Really good point. Evolving a system gradually while maintaining a clear architectural direction seems much safer than attempting a complete rewrite. The tricky part is knowing when the old foundation has become too expensive to maintain.
A question prompted by the recent FAA ground stop in the US.
8 Comments
I'd start slicing where a failure is cheapest to notice, and prove each slice against the old system before it takes live traffic. Run the new slice in shadow on the same production inputs and diff its outputs against the old one for a few weeks. Every mismatch is either a bug or one of those undocumented behaviours you mention, and you want to learn which while the old system is still the one answering.
The FAA's January 2023 NOTAM outage shows where the risk sits during that period. The FAA said contractors unintentionally deleted files while working to correct synchronisation between the live primary database and the backup, and the nationwide ground stop followed. So the seams between old and new, and the sync jobs that keep them aligned, need the same tests and change review as the systems on either side of them.
Please log in to add a comment.
It should probably be a fundamental law of software by now that the only way to build a successful system is to start small and evolve it. When you have a "heritage application" that contains decades of knowledge, a big band rewrite is a guaranteed disaster. Instead, you should start identifying seams that allow you to split it, so you can work on small pieces to improve or replace them.
People often refer to other engineering disciplines and say "they can specify a whole building up front, with all the details captured in blueprints and architectural diagrams", but building software is less like constructing a building and more like planning a city... not to mention what happens "on the ground" when construction workers apply their mētis to perform "anarchistic tasks" that make a bad plan succeed in reality.
Please log in to add a comment.
Thanks for the thoughtful comments.
@Sitesh Pradhan raises an important question: when does the cost of preserving an existing foundation outweigh the value it provides?
@Mike Dabydeen brings in an operational perspective. Shadow traffic, comparing outputs and validating the seams between old and new systems are practical ways to discover undocumented behaviour before making the switch.
@Steve Fenton's city-planning analogy reminds me of Christopher Alexander's work, particularly "The Timeless Way of Building" and "A Pattern Language".
Alexander argued that successful environments emerge through incremental, context-sensitive changes, guided by a shared language of patterns. The whole evolves through many smaller acts of building, rather than being fully specified upfront.
@Dilip Kumar brings us back to the importance of design.
Taken together, these perspectives suggest an interesting balance. We need architectural direction, but also the humility to recognise that we cannot anticipate everything upfront.
Maybe good software architecture is less about specifying everything we intend to build, and more about creating the conditions in which a system can evolve while retaining its coherence.
Please log in to add a comment.
Please log in to comment on this post.
More Posts
- © 2026 Coder Legion
- Feedback / Bug
- Privacy
- About Us
- Contacts
- You Tube
- Tiktok
- Premium Subscription
- Terms of Service
- Early Builders
Related Jobs
- Tier-2 Ground Terminal / IT Technician _ Junior, Intermediate and Senior Levels TS/SCI with PolygraphGeneral Dynamics Information Technology · Full time · Australia
- Tier-2 Ground Terminal / IT Technician (Australia) Junior, Intermediate and Senior Levels TS/SCI with PolygraphGeneral Dynamics Information Technology · Full time · Australia
- Tier-2 Ground Terminal / IT Technician (Australia) Junior, Intermediate and Senior Levels TS/SCI with PolygraphGeneral Dynamics Information Technology · Full time · Australia
Commenters (This Week)
Contribute meaningful comments to climb the leaderboard and earn badges!