OODBs appeared when the problems to be solved using OODBs accumulated.

Up to some time, you had rather un-complex, straightforward, although sometimes large data structures, on which you needed to perform pretty simple operations, so that each record could be processed mostly the same.

As OO became mainstream, more complex processing became possible, applications became smarter, and people started to expect more from their data stores. Which created pressure on programmers, which constantly had to deal with the impedance mismatch between OOP and RDBMs.

Two solutions emerged: ORM solutions, which tried to provide a generic workaround for the impedance mismatch, and OODBs, which tried to eliminate the mismatch altogether. Problem is, only few problems had only a part which was perfectly solvable by using OODBs, while for most problems just a small part was supported by OODBs, whereas most of the data was still ideally handled in a relational way. That’s why ORM solutions still thrive, whereas OODBs aren’t used that much.

Then came Google. Google wants to index everything. They don’t care if it’s a picture, a document or whatever, they want you to be able to search for it. How does this fit into the RDBMS concept? I’d say pretty bad – what do a JPG picture of a dog and a GIF containing an UML diagram have in common? This doesn’t even fit into the OO concept, since you can barely identify classes, at least not at the source code level – your application has to do the classification for you. Plus, your data model is huge, with most RDBMSs or OODBs not being able to support that amount of data. Which is why they invented BigTable – essentially a container into which you can write stuff, very much the same way you write bulleted lists while brainstorming.

Is this a new attack on RDBMSs? I wouldn’t say so – it’s simply a new solution to a new problem, one which RDBMSs are ill equipped to solve. Of course, Google could have created a new mapping library which should just store any BigTable inside a large cluster of RDBMS servers. But what’s the point? You don’t get efficiency by shoehorning a data structure on a storage system that doesn’t support it well.

I do not think I will go to the extent of up/downloading the database content raw because I do need to process the data on the server (which at the moment is Google because it’s CHEAP) but I do like the idea that databases are nowadays ubiquitous and make our life easier.

Thanks for this laid back, pertinent article !