Why I don’t like the term “NoSQL”

This is a rant post, but just to clarify things, it’s not a rant against the use of non-relational databases. I think that the shift in recent years from a world in which relational databases are used almost exclusively regardless of what the need is, to today’s situation where it is possible and even considered a good idea to choose the best fitting solution from any number of data storage paradigms, is a truly blessed change. I am a big fan of some non-relational database solutions, and to be honest as a programmer I enjoy using some of them more than I enjoy MySQL or any other relational database.

This is a rant against the too-common term “NoSQL”. In my opinion, “NoSQL” is an example of layman terminology which does not properly describe the concepts which in most cases it aims to describe, and should not be used by professionals which are technical enough to understand the true meaning of these concepts.

“NoSQL” databases are all about the data model – in most cases, the term is used to describe any kind of storage engine (or database) in which data is stored in non-relational manner: object storage, document storage, key-value storage etc. Indeed, the term is more about what the database is not that about what it is.

Relational data is data that can be described as a table – in contrast to what some think, the term “relational database” has nothing to do with the ability to define and enforce relationships between data in different tables. If this was the case, MySQL using the MyISAM storage engine would not be a relational database. The term “relation” is a mathematical term, which existed before the creation of relational databases and is used to describe a relationship between two finite data sets, which can be described in a tabular manner (and I am not a mathematician, not even close – so I apologize in advance for this likely inaccurate description).

But, SQL has nothing to do with this – SQL is the language used to send commands to the database, and nothing more. It is true that there is an almost 1-to-1 correlation between database engines that store data in a relational manner and database engines that use SQL as a query language, but saying that relational databases are SQL databases is like saying that  (and assume it’s 1984 again) the Russian language should be abolished when in fact we want to say that communism is an unfitting economic system. It’s a poor way to describe your intentions, and it makes you sound like an ignorant moron.

There are many client libraries and wrappers that allow you to query a relational database such as MySQL and Oracle without writing any SQL code yourself. This doesn’t make them NoSQL databases. Some popular non-relational databases, such as Amazon SimpleDB and the Google App Engine Data Store provide query languages that are quite similar to SQL. This doesn’t make them SQL databases.SQL is just a language, and it is a good one for what it’s supposed to do (putting aside all sorts of discrepancies between vendor-specific SQL implementations). SQL is not what NoSQL databases are NOT about.

So, next time when you want to use a term that describes all databases that do not store data in a tabular manner, use the term “non-relational” or if you really like acronyms, “NonRDBMS”, and not “NoSQL”. Or even better – use a term that describes what your preferred solution is, not what it is not. After all, when you say “non-relational storage engine”, you are probably not referring to your file system, right?

One thought on “Why I don’t like the term “NoSQL”

  1. Relational data is not equal to table. Relational data model and relational algebra that stands behind it also about relations between entities (read rows in same/different tables).

    Yes MyISAM has no foreign keys support (responsibility on relational consistency delegated to application) – that doesn’t make it non RDBMS. It still uses all the power of relational model to make joins, for example.

    And SQL (ANSI) is not just a language – it’s a (sub)dialect of relational algebra and is strongly connected to relational DBs.

    NoSQL is a bad terminology, agree completely.
    But paragraphs about relational data and SQL are just wrong, sorry.