News for hackers. Stuff that matters.
Re:Vector database is not a separate database category by Guvante on Tuesday October 03, 2023 @03:29AM (#37748895)
I mean NOSQL was hype with no substance but "you can scale more if you deal with not having ACID" is just generally true.

Of course ACID scales to well into the Fortune 500 scale so...

Re:Vector database is not a separate database category by smitty1e on Tuesday October 03, 2023 @05:22AM (#37749591)
"No substance" seems a bit harsh.

They mostly seem a tarted up associative array, sure, but a key-value store is a thing.

Re:Vector database is not a separate database category by mnky9800n on Tuesday October 03, 2023 @05:48AM (#37749743)
But don't you prefer your key value stores to be wearing red lipstick and a pushup bra?
Re:Vector database is not a separate database category by threeseed on Tuesday October 03, 2023 @06:25AM (#37749984)
> but a key-value store is a thing

DynamoDB underpins much of AWS which in turn underpins a ridiculous number of web services.

So definitely more than just a thing.

Re:Vector database is not a separate database category by rakoo on Tuesday October 03, 2023 @08:31AM (#37750985)
A naturally distributed key/value store. NoSQL wasn't a product, it was a radical rethinking of the balance between what we wanted and what we needed. Turns out some absolute necessities of using a RDBMS or even using SQL are not that important, and relaxing those actually-not-requirements allows massive scalability, something we desperately needed, or some evolved data structures and computation, like what redis provides.
Re:Vector database is not a separate database category by b33j0r on Tuesday October 03, 2023 @05:25AM (#37749611)
Hype with no substance? Hold on ;)

I know it hurts, but remember that we were still writing XML way too often in 2013. Document stores were purpose-built for a thing we needed. And the execs bought in, to boot.

At the time, it was very convenient to shove JSON and similar in an indexable and aggregate-able database. I implemented a lot of very very large-scale code analytics via mongodb--about 10 years ago.

There wasn't really support for unstructured types in mysql or postgres or oracle without some additional transformation pipelines or db extensions back then. We usually just wrote stupidly complicated SQL schemas that took up our whole conference room, and only that one stressed-out guy really understood.

So it was neat. I would choose something better today.