3 ms·
Potential Consistency
- cperciva 16y agoSummary: Systems designed to be consistent are bad, because if they break, they will be broken. At least, that's what I'm parsing this as. The author argues that because eventually consistent systems are designed to have inconsistency as a normal event, they can recover from inconsistency -- but this completely ignores the fact that large consistent systems (or at least, those which are designed well) also check for inconsistencies so that fail(ing|ed) nodes can be killed. Maybe I'm missing something: Did anyone see something profound in this article?