Part 4 of 8 · 1 min read
Which seam first
Choose by benefit, not by neatness.
Splitting for its own sake wastes effort. BM gives five reasons to pick one seam before another.
| Driver | Reasoning |
|---|---|
| Pace of change | A big batch of inventory changes is coming, so make the warehouse separate to move faster |
| Team structure | One team is in London and one in Hawaii. Split off what the Hawaii team works on most so it can own it (page 5) |
| Security | An audit says to protect sensitive data more. Finance code that handles it becomes its own service with extra protection |
| Technology | A team wants to try a new algorithm in another language. A separate recommendation service makes a second version easy to test |
| Tangled dependencies | Pull out first the seam that the least code depends on |
For the last point, BM suggests viewing the seams as a graph of dependencies. A seam on which many others depend is hard to remove. The hardest tangle is usually the database, covered next.
Read it as
Ask what the split will make easier. If the answer is not a change that is coming, a team, a risk, a technology or a low-cost cut, wait.
Check yourself: 1. A seam is used by nine others. Is it a good first choice? Why? 2. Which driver suits "our payment code is audited every year"? 3. Why does team structure appear among technical reasons?