It needs to be designed in, as a bespoke goal, from the start. And I think that's doubly true for #glowdust, which is built with #asmalldbms principles, where usability is fundamental.
Which brings me to my current work. I think that, even though the current json based store for glowdust is "experimental", it must work with future versions, which will have more efficient file formats and different execution engines.
Which is very, very hard. But that's architecture for you.
2/2
@rotnroll666 I hate "The Cloud(TM)", its providers and all the software that is being built to run exclusively on it.
This trend fundamentally moves power away from the hands of the user and stifles innovation and creativity.
That's precisely why I wrote my #asmalldbms post and why I am working on #glowdust with real hardware in mind.
Not only do we not need all this complexity and waste, it's actively harming everyone.
You may have read my post about big data not really being a thing:
https://radiki.dev/posts/a-small-dbms/#commodity-hardware-is-all-you-need
but you may have remained skeptical. In that case, here's someone else saying the exact same thing:
https://motherduck.com/blog/big-data-is-dead/
Marketing angle aside, the conclusions are the same. Big data is not real and even if it was, it's not useful.
@Anachron Yes.
That's core to my #asmalldbms idea. I want a dbms that is built to be small and useless for data capitalists.
I want an Internet of communities, where wealth is created by people interacting with each other, not by hoarding data.
I don't know how to build this, but I have some ideas. Or maybe questions? I put them in a blog post:
https://radiki.dev/posts/a-small-dbms/
Please boost. This is the first time I ask, but this subject is very dear to me and I want to reach a wide audience.
Before large model ML techniques got their recent day in the sun, the main way to monetize large data sets was by mining the social graph, the metadata in the data. We know this as surveillance capitalism.
AI instead tries to monetize the actual data _and_ sweep any ownership issues under the rug.
This will be known as the age of (overt) data capitalism.
[...] a massively decentralized Internet, consisting of a huge number of small communities, requires database management systems that are built from the ground up for the purpose.
The systems available today are built to support massive, centralized, owned datasets, at immense complexity and cost, with intentions that go against the principles of a free and open Internet.
"In a winner-takes-all environment, it is inevitable that every DBMS vendor will claim that their implementation, model or language is the one true way of managing data. This is in tension with standardization processes at the language level, which promises, but barely delivers, some degree of portability. The end result is the same as for everything capitalism touches. A race to grab territory and dig trenches around it, at the expense of innovation."
And what does that punk DBMS look like?
[...]
- It must be optimized for real hardware and self hosting, going against the consolidation attempts of huge corporations to control every bit of the software stack.
[...]
What would be the political message of a new DBMS?
I argue it wouldn't make sense to be anything but countercultural. Software is by default the result of capitalist exploitation and a tool for creating profit. That is the dominant narrative in the software industry, and criticising it requires taking the opposite position.
Chris Gioran 💔