2024-09-18 Architectural Decision-Making Processes

We know that a project of this size is faced with architectural decisions of various sizes, ranging from things scoped very narrowly, which impact only one or a few modules/teams, to platform-wide decisions impacting almost everyone.  A one-size-fits all approach to making these decisions is unlikely to work well, and we've seen this already with the RFC process.  Let's enumerate the types/sizes of decisions we make, then try to identify the applicable processes for each.  If there are gaps, do we need to define additional processes or guidance for dev teams, and architects?

Previous Notes:

Types/sizes of technical decisions

• Team/Module-specific 
  • Process:  ad-hoc/team-specific, sometimes with the help of SAs
  • Who:  typically make internally by dev teams
  • Example(s): 
    • Whether certain functionality is added to an existing module, or to a new module.
    • Technology choices - shared libraries, frameworks, etc., e.g. quartz timers, various npm packages
    • Storage layout, e.g. jsonb vs traditional relational tables, etc.
• Problem-specific
  • Process:
  • Who:
  • Example(s):
    • Pub-sub/Event-driven communications 
    • Adoption of spring-way
• Security-driven
  • Process:
  • Who:  Security Team
  • Example(s):
    • ...
• Platform-level/Fundamental
  • Process:
  • Who: 
  • Example(s):
    • Eureka
    • Consortia support (ECS) and cross-tenant requests (possibly problem-specific)
• UI architectural decisions
  • Process: A proposal or problem is brought to stripes-architecture, it's discussed, and a decision is made.
  • Who: Attendees of the stripes-architecture meeting
  • Example(s):
    • How GitHub actions are organized/implemented/used
    • Migration away from Yarn v1

Today:

• ...