We Don't Need a "Database" | Hacker News Reader

We Don't Need a "Database"

rashkovskii.com

15 pointsdavidkellis10 comments

Written by a programmer who has a myopic view of how data is used. Data is not used by just one application in the real world over time; if your app that uses the stupid database becomes popular (or your startup becomes popular), you will now need many interfaces into that database - reporting, analysis, OLTP, maintenance. By putting it in a "persistence-aware programming environment" you're *&%$ing everyone else except you. Woohoo!

With a "persistence-aware programming environment" that uses a database as the backend, you have both. So I'm not sure if it's valid to argue PAPE VS DB...

This problem applies to using any new or non-standard technology without abundant interfaces to a wide variety of tools. That's the downside. The advantage is that you get to use something better, and maybe it makes you more productive, and maybe it catches on and advances the state of the art.

If everyone always used whatever interfaced best with the rest of the world, new technologies would be adopted much more slowly.

I was not arguing about "prohibiting" third-parties to connect to your database.

From the article: "What we really need is a persistence-aware programming environment"

This really hits home with me. I've been working on a data model for one of my projects. I'm trying to model time series data so that I can perform analysis of the data as it changes over time. I've been thinking about how best to model this using the relational model, but even after a few months I've been unable to develop a relational model that I'm happy with.

The more I read about MapReduce and similar methods of computation, the more I become convinced that I could model time series data and perform analysis of it with a method of computation that is more "integrated" into my data store. Essentially, I'd like to be able to use my programming language of choice (at the moment, that's ruby) to perform queries on my datastore that for one reason or another I've been unable to represent in SQL or a similar query language.

Agreed. ObPlug: if you're coming to RailsConf, stop by my talk on MagLev (http://ruby.gemstone.com), which could certainly be described as "a persistence-aware programming environment" for Ruby.

I wish I was able to afford to come to RailsConf, I was interested to visit it to be able to chat about MagLev and Rubinius specifically

Have you thought about a dimensional model? IE a star scheme with a standard time/date dimension and your facts.

"We have persistent objects, they're called files." -- Ken Thompson.

With a 4KB minimum block size? That should do nicely for all of my 100 byte postal addresses; NOT.

Reply on news.ycombinator.com