Aviation isn't perfect; nothing implemented by large groups of fallible humans with budget constraints will be. But it has one of the best safety track records of any industry.
Now, why won't aviation style engineering be applied in other fields, like databases? Well, because no one really cares enough. No one dies if some random database used by some ad platform loses the occasional transaction. Yeah, it's frustrating to engineers who are trying to build reliable systems, but in the grand scheme of things losing a few percent of transactions isn't the end of the world for most businesses.
You get safety cultures like that in aviation because there are real, substantial risks, so you need to have thorough engineering discipline, properly designed redundancy, etc.
For databases that are used for the majority of the business world, efficiency is generally a bigger concern than correctness; they'd rather have cheap and fast databases that lose a few transactions occasionally than something that actually provides consistency. But of course everyone thinks they need consistency, so it's advertised as a selling point while not actually being provided in practice.
> For databases that are used for the majority of the business world, efficiency is generally a bigger concern than correctness; they'd rather have cheap and fast databases that lose a few transactions occasionally than something that actually provides consistency
'majority' is a strong claim. Any figures to back it up?
No, no figures to back it up. This is just based on anecdotal experience, having never seen very many businesses that actually prioritize consistency or testing or validating it. The fact that Kyle keeps on finding these massive problems in popular databases is part of that anecdotal evidence.
Obviously people want correctness; they would like their databases to not randomly lose data. Hence the fact that it's a highly advertised feature.
But when it actually comes down to selecting a database, convenience and performance seem to be what people actually compare on, there are very few places that actually hire someone like Kyle to dig in and verify the the consistency claims about a database.
I'm positive people want their databases to be correct. There are databases which promise great speed in return for occasionally losing bits of your data, and they get very little use outside of special uses.
Exactly; people would like their databases to be correct and consistent, but generally don't care enough to actually do something like hiring someone like Kyle to verify it before buying a database. They just find something with the right features and performance, and go ahead and use it. You see a lot of them in this thread; people who used RavenDB for a while, and then had to migrate away because of issues. If the business actually cared enough about consistency, there would have been some kind of validation and verification before selecting a database.
