I already don't trust anything a bot can POST to. So now I'm working towards making all my personal information live on servers that I control.<p>And when it comes to decentralized personal information, it's rare that I (anecdotally) have enough of anything to require a true DB. For instance:<p>- All of my health records (excluding imaging) could easily fit neatly in a 2MB JSON.<p>- A list of every photo I've ever taken, a file path to each, and a description would maybe take 10MB.<p>- A description of every event I've ever been to, and every person I've ever met might take 8MB.<p>- A lat, lng, and timestamp entry every minute for 10 years would be 5,256,000 entries × 26 bytes = ~130MB<p>Given how little overhead it takes to store the entire information of any individual, plus how infrequently it is likely to be modified, it seems there is no good reason to use databases in a decentralized context.<p>What do you think of this prediction?
I’ve thought about this a lot before in my personal and professional life. If you combine JSON with a SQLite database, you can get the best of both worlds.<p>Disclaimer, I wrote this for work but it is based on ideas I developed outside of work. <a href="https://pinata.cloud/blog/will-json-work-as-a-database/" rel="nofollow">https://pinata.cloud/blog/will-json-work-as-a-database/</a>
Ignoring the prediction that everything will be decentralized in the near future (it probably won't), how about this scenario: you want to see if the voting activity (recorded in your events) correlates with the number of times you went to the doctors in the year prior to election (as recorded in your health records). If we want to be able to run such a distributed query, we need to be sure that each node stores the data in predefined format ("tables") and that one of their services run a service that will receive the distributed query request, decide whether or not it wants to participate, and then execute the business logic of the conditions (healthevent.year > getdate() - 365), probably defined in some programming language ("SQL").
From your description all would easily fit into RAM and full-scan searches across all records will be fast enough. Browsers hardly sweat reading 200MB data files. No need for indexes, optimized storage, compression or special query languages. Unless you'd worry about scale (managing the same for 1000 people). It's a luxury of modern hardware.
I think databases or similar data stores will remain, but for many projects looking at performance or small memory footprint will no longer matter. I've used Postgresql for tiny projects (all in docker containers) because it was convenient.
> What do you think of this prediction?<p>That you don't know what you are talking about. You might want to reconsider what you think a database is.