Most architecture talks are about what to choose. This one was about which choices are expensive to reverse — data modelling, service boundaries and identity — and how to defer the rest until you know more.
At a glance
1 day
six sessions, one workshop
120
engineers and architects
4
case studies from production systems
The agenda
Sessions ran back to back with a break between each block.
09:30 · Opening
Which architectural decisions are genuinely hard to reverse, and which only feel that way.
10:30 · Data modelling
Getting the boundaries right before the schema hardens around the wrong ones.
12:00 · Service boundaries
When splitting a monolith helps, and the far more common case where it does not.
14:00 · Identity and access
Designing permissions that still make sense after three reorganisations.
15:30 · Workshop
Attendees mapped one of their own systems and argued about it in small groups.
17:00 · Close
Open questions and what people were taking back to their teams.
What people took away
- Defer the decisions you can; commit early only where reversal is cheap
- A service boundary drawn around a team rarely matches one drawn around data
- Permissions modelled as roles age badly — model them as capabilities
- Every architecture decision deserves a written record of why, not just what
The most repeated comment afterwards was that the useful part was hearing which decisions other teams regretted.
See what we do
Practical questions
The talks were; the workshop was not, so people could speak freely about their own systems. Get in touch if you would like the recordings.
Two or three times a year, usually alternating between engineering and product themes.
Want to hear about the next one?
We run these two or three times a year. Tell us what you would want covered.
