There is some FUD in this blogpost as well as in the comments in this thread.
I think the current status quo for databases is canned software. And this isn't necessarily bad because neither of the three databases mentioned hide their specs or default settings, the three have very good docs and community willing to help, in addition to companies giving commercial support. Whats your excuse to misuse these products?
RethinkDB writes your data to disk before acknowledging the write but on the other hand can't elect a new primary in case of failure, two completely different features/limitations that might work for someone and not for other ones. Is that hard to understand? Did mongo documentation lie you at some point?
Accept that you are "buying" a general purpose product, the designers thought that their users will need those features, deal with it.
Otherwise build your own database, I know this might sound very hard but I guess in the future we will see smaller building blocks that let you build something that handle your needs like this:
I was just thinking about this issue in the last few days. I'm working on a side project, due to the nature of the data model, converting back and forth to fit into a relational database is kind of annoying, so I was looking around for other databases.
Right now, the situation with database is that we have to convert our internal data structure into a representation that fit the data model of the database we're using (ie rows for relational, document/key for the NoSQL group). I can see the reason the data model has to be that way for scaling, distributed etc... But if I'm happy to scale my database up, and would prefer to have the database storing the data as close to the memory data structure as possible (similar to object databases -- albeit with a boarder definition of "object"), is there any database that could do that?
Otherwise, is there any suggestion on how I could get started to build one?
Which is also my point, both mongodb and rethinkdb has the same feature in this regards but different defaults. Does that makes one worse than the other one?
Yes, because lots of people end up using software with default values. Can you blame them? Perhaps... in a way that you can blame a user who accidentally clicks a "delete account" button that doesn't have a confirmation step.
Good defaults are expected in quality software, and are just as important as any other part of software interface, CLI or GUI.
I think the current status quo for databases is canned software. And this isn't necessarily bad because neither of the three databases mentioned hide their specs or default settings, the three have very good docs and community willing to help, in addition to companies giving commercial support. Whats your excuse to misuse these products?
RethinkDB writes your data to disk before acknowledging the write but on the other hand can't elect a new primary in case of failure, two completely different features/limitations that might work for someone and not for other ones. Is that hard to understand? Did mongo documentation lie you at some point?
Accept that you are "buying" a general purpose product, the designers thought that their users will need those features, deal with it.
Otherwise build your own database, I know this might sound very hard but I guess in the future we will see smaller building blocks that let you build something that handle your needs like this:
https://github.com/rvagg/node-levelup/wiki/Modules#plugins