Sorry for digging this up, but as we just discussed the “WW3 assumption” in today’s randomness summit I wanted to share my thoughts.
I feel the CAP theorem should not be used as guidance for designing complex blockchain systems. The theorem just lacks strict assumptions, for example regarding latency and is plainly: confusing. Which is why there has been criticism.
Back to the topic though:
Why would you even prefer liveness/availability over consistency?
If Ethereum was mainly a gaming platform and you wanted to accept the possibility of losing a magic sword because of a fork, then okay, favor availability. But Ethereum is currently being utilized to issue bonds worth millions of Euros and the recent DeFi craze implicates how important consistency is. If you consider a fork ending up in two separate networks as “healed”, what really is won here? Did my stable-coin assets simply double a few years after the war when the Internet fully came back?