Notes on Distributed Systems for Young Bloods (2013)

I believe he's suggesting that caches should only be allowed to read from persistent storage and that values read from cache should not be assumed to be current. For example, if I read $FirstName, $LastName from some cache lookup for a user and then go to update $Address I should not write $FirstName and $LastName along with it as a "whole user update" sort of thing; just update the value that you need to.

I think this is referring to write-back caching, where the write operation is confirmed when the cache accepts it, but a separate process is responsible for replicating the write operation from cache to the persistent storage.

Write-through is more widely accepted, since the operation is not confirmed until it's successful on both the cache and the persistent storage.

Also this:

"Gotta get this off my chest. While "youngbloods" is more common, it felt too gendered, too dudebro. "Young bloods" was my compromise. There."

https://twitter.com/jmhodges/status/423315878722031616