The Memcached Rehabilitation and Redis Skepticism

A post praising memcached sparked a real debate about whether Redis is being overused as a cache. The core argument: Redis is powerful but comes with footguns when used as a cache specifically, requiring wrapper libraries and careful TTL discipline to avoid it silently behaving like permanent storage. Memcached, being a pure cache with no persistence, can't be misconfigured in the same way.

One commenter described a specific operational practice: wrapping the Redis client library so it's impossible to store anything without a TTL. That's a real engineering discipline that most teams skip, and it illustrates the gap between Redis-as-designed and Redis-in-practice.

The deeper pattern: there's a recurring HN theme of 'boring technology' advocates pushing back against the default choice of whatever is most feature-rich. Memcached, SQLite, and Postgres keep winning rehabilitation threads. The counterpoint raised in the comments is that Redis does many things beyond caching, and for teams that need pub/sub or Lua scripting, the complexity is justified.