If you are not a NoSQL supporter, then you are interested in none of the bulletted items.

-Todd

I find that many complaints about relational theory are there because of lack of understanding of the ceoncepts and of SQL, and exposure to poorly constructed and performing databases.

The other complaints related to hierarchical constructs, networks, objects, partitioned high scale operation, etc. are all valid. Since I focus more on the analysis side, I prefer “not only” where many of my peers working on web applications are doing form/transaction style work and “no” is a better fit.

The problem with classifying all these information storage technologies and models is similar to the problem early libraries had (and still have) regarding how to organize and structure their collections. In the end, I prefer Cutter’s solution over Dewey’s because an attribute model is more flexible than a fixed taxonomy. You’re applying the fixed taxonomy model to the databases. I think an attribute model might be better suited because then you can map it to the type of use or specific requirements needed by different people*

The problem is that attribute-based organization doesn’t lend itself to simple lists or tables of “is-a this”, e.g. MarkLogic is a document store and XML database.

*Best argument for not having mutual exclusion in product classification is the floor wax – desert topping debate.

I find that many complaints about relational theory are there because of lack of understanding of the ceoncepts and of SQL, and exposure to poorly constructed and performing databases.

The other complaints related to hierarchical constructs, networks, objects, partitioned high scale operation, etc. are all valid. Since I focus more on the analysis side, I prefer “not only” where many of my peers working on web applications are doing form/transaction style work and “no” is a better fit.

The problem with classifying all these information storage technologies and models is similar to the problem early libraries had (and still have) regarding how to organize and structure their collections. In the end, I prefer Cutter’s solution over Dewey’s because an attribute model is more flexible than a fixed taxonomy. You’re applying the fixed taxonomy model to the databases. I think an attribute model might be better suited because then you can map it to the type of use or specific requirements needed by different people*

The problem is that attribute-based organization doesn’t lend itself to simple lists or tables of “is-a this”, e.g. MarkLogic is a document store and XML database.

*Best argument for not having mutual exclusion in product classification is the floor wax – desert topping debate.

Cheers, Tony.

thanks for your posting and your ideas. I really do appreciate your opinion. Nevertheless I allow myself to write down the rationale behind my decision for the world and for a fruitful discussion.

(I included your blog to the NoSQL archive)

You adressed some points from my NoSQL-database.org website:

> Who decided that being open source was a requirement to be real NoSQL system?
No one decided this. I didn’t wrote this. What I wrote was: “…mostly address some of the points”. So it’s clear that I mean a good mix
of all points makes it NoSQL.

Then I distinguish between “core” and “soft” NoSQL at the nosql archive. First of all, I was very glad that other non relational databases joined the NoSQL movement. And I was happy to include them on the website. I received several hundred emails and it was great to see how people loved to see the growing space. And get the freedom of choice back. Even people from the surroundings of Oracle and IBM emailed me to join the boat. And no system was left excluded.

I follow the NoSQL movement since early 2009 after some talk with Jann Lennart (CouchDB), who worked not far from my home. Perhaps this is why I personally feel that the reader of my website should see that the inventors of the “NoSQL” word had a very specific area of databases in mind. If you recall the history the new term NoSQL, it was “invented” by Johan Oskarsson (I know that the ‘true’ history goes back to the 1970s…) and first blogged by Eric Evans. They set up conference for a specific group of databases (NoSQL) designed mostly to back specific websites. So if was not about being “cool and they want to be like Twitter, Google, and Facebook” as you mentioned. It was a conference about real databases like HBase, CouchDB, Cassandra etc. and for websites with real needs and people interested in this.

And the readers of nosql-database.org should see, that there are concrete ‘core’ databases these people had in mind. And of course that there are other great non-relational databases which support NoSQL too (my soft part).

Perhaps the word “intention” or “soft” I have chosen in this sentence “soft nosql: not the original intention of ‘NoSQL’ …” is not perfect. Do you have a better proposal? I promise I think seriously about it.

And the sentence above I wrote continues (you left this out in your citation): “…worth a look for great non relational solutions”. From this you can easily see that I highly appreciate other great databases as XML-, Object Databases or others in the broad sense of NoSQL.

Best Regards and good luck to MarkLogic
Stefan Edlich

It’s purpose might be to describe documents. Talking about XML databases makes little sense to me, from a technical standpoint the important part is that the underlining data model allows you to store tree-like structures.

From a business perspective what’s important is the fact that you can store your business objects in a model that is suited to fast retrieval and insertion of those objects.

So drop the XML label, drop the JSON label. They can both describe the same (on a side discussion I would say JSON is better for describing objects and XML for annotations and documents). But what really matters is what can be achieved with these models that cannot with a simple table. The shift from data to information.

My 2 cents – by the way great article!