add a specifications page for the provider cache (#656)

Signed-off-by: Thomas Poignant <[email protected]>

website/src/pages/specification/provider-cache.md

@@ -0,0 +1,32 @@
+---
+title: Provider Cache Specification
+---
+
+# Specification Document for OpenFeature Providers Cache
+
+|                      |                 |
+|----------------------|-----------------|
+| **Creation Date**    | 06/04/2023      |
+| **Last Update Date** | 06/04/2023      |
+| **Authors**          | Thomas Poignant |
+
+## Overview
+
+OpenFeature Providers are used to call the GO Feature Flag the relay proxy to evaluate flags.
+The purpose of this specification document is to outline the requirements for implementing the cache policy in the 
+providers that is compatible with GO Feature Flag.
+
+## Requirements
+
+1. The provider should implement an LRU cache with a configurable size and TTL.
+2. The cache size and TTL should be configurable by the user of the provider.
+3. The cache should be activated by default, but it should be possible to deactivate it in the configuration.
+4. The cache key should be a combination of the flag name and user key.
+5. When a flag is evaluated, the provider should first check the cache for the result before querying the relay proxy.
+6. When there is no entry in the cache, we should query the relay proxy and update the cache with this entry.
+7. When we have reached the max time for a cache entry, it should not be possible to retrieve it.
+8. When we have reached the max size of the cache we should remove the oldest entry.
+9. We should collect all the flags usage and call the API `/v1/data/collector` for statistics reason.
+10. The data collection should be activated by default, but it should be possible to deactivate it in the configuration.
+11. We should not call the relay proxy API `/v1/data/collector` for every evaluation, but call it in bulk with a periodic threshold.
+12. The threshold should be configurable by the user of the provider.