4 ms·
I can agree to an extent, but I think a lot of this can be mitigated by having tight, isolated connections between parts of a system where each result can be va
by jjice 2y ago
I can agree to an extent, but I think a lot of this can be mitigated by having tight, isolated connections between parts of a system where each result can be validated. Then you can address each step/service in isolation based on the inputs and outputs of each. This is sometimes harder to achieve, but having specific inputs and outputs at each step makes it easier to identify which step is "faulty". It's not always perfect, for example in the case of a validation library that is erroneously allowing incorrect data though, but you can hopefully identify the surrounding steps and narrow your search.
- coliveira 2y agoThis is the theory, and it can work if done properly. However, just like in a GOTO-ridden language, modern distributed systems are extremely difficult to setup in this optimal way. The more components we have in our distributed system, the harder it becomes to have any sense of safety and correctness. No wonder we started to see systems imploding worldwide as it happened last week.
- lesuorac 2y agoWhile connecting say two systems of 99% reliability will be better than say connecting two systems of 50% reliability I don't think that's the real issue. I find the real issue is that there is no error propagation nor audit trail. Somebodys system notices a problem and does like `throw new IllegalStateException("Invalid Arg")` and good luck figuring out what happened. A lot of system/code aren't designed to be tested easily and they just accumulate bugs because nobody can fix them.