Normalization is from the devil

time to read 4 min | 723 words

The title of this post is a translation of an Arabic saying that my father quoted me throughout my childhood.

I have been teaching my NHibernate course these past few days, and I had come to realize that my approach for designing RDBMS based applications has gone a drastic change recently. I think that the difference in my view was brought home when I started getting angry about this model:


I mean, it is pretty much a classic, isn’t it? But what really annoyed me was that all I had to do was look at this and know just how badly this is going to end up as when someone is going to try to show an order with its details. We are going to have, at least initially, 3 + N(order lines) queries. And even though this is a classic model, loading it efficiently is actually not that trivial. I actually used this model to show several different ways of eager loading. And remember, this model is actually a highly simplified representation of what you’ll need in real projects.

I then came up with a model that I felt was much more palatable to me:


And looking at it, I had an interesting thought. My problem with the model started because I got annoyed by how many tables were involved in dealing with “Show Order”, but the end result also reminded me of something, Root Aggregates in DDDs. Now, since my newfound sensitivity about this has been based on my experiences with RavenDB, I found it amusing that I explicitly modeled documents in RavenDB after Root Aggregates in DDD, then went the other way (reducing queries –> Root Aggregates) with modeling in RDBMS).

The interesting part is that once you start thinking like this, you end up with a lot of additional reasons why you actually want that. (If the product price changed, it doesn’t affect the order, for example).

If you think about it, normalization in RDBMS had such a major role because storage was expensive. It made sense to try to optimize this with normalization. In essence, normalization is compressing the data, by taking the repeated patterns and substituting them with a marker. There is also another issue, when normalization came out, the applications being being were far different than the type of applications we build today. In terms of number of users, time that you had to process a single request, concurrent requests, amount of data that you had to deal with, etc.

Under those circumstances, it actually made sense to trade off read speed for storage. In today’s world? I don’t think that it hold as much.

The other major benefit of normalization, which took extra emphasis when the reduction in storage became less important as HD sizes grew, is that when you state a fact only once, you can modify it only once.

Except… there is a large set of scenarios where you don’t want to do that. Take invoices as a good example. In the case of the order model above, if you changed the product name from “Thingamajig” to “Foomiester”, that is going to be mighty confusing for me when I look at that order and have no idea what that thing was. What about the name of the customer? Think about the scenarios in which someone changes their name (marriage is most common one, probably). If a woman orders a book under her maiden name, then changes her name after she married, what is supposed to show on the order when it is displayed? If it is the new name, that person didn’t exist at the time of the order.

Obviously, there are counter examples, which I am sure the comments will be quick to point out.

But it does bear thinking about, and my default instinct to apply 3rd normal form has been muted once I realized this. I now have a whole set of additional questions that i ask about every piece of information that I deal with.