Lazy loading of ruleset: Implement caching layer for Spectral ruleset to avoid reloading on each validation request. Currently the entire ruleset is reloaded per request - consider using an in-memory cache with TTL or file-watch invalidation strategy. This will scale better as the number of custom rules and functions grows.

Conditions of acceptance

• Decide on the best cacheing strategy (in memory, in MediaWiki, etc).
• Establish baseline and capture updated performance metrics.

Implementation details

The actual mechanism might be different from the spec cacheing. This is a fast operation that happens multiple times within the same process; spec caching is more focused on maintaining a cache across processes.

Potential people to talk to for input, recommendations, code review:

• Release Engineering (Tyler, Reedy, Hashar)