The history of transactional relational database systems is one entwined in attempts by people from the IT world to classify things in an unambiguous and structured manner. Underpinning the principals of relational database theory is the idea that a piece of data should exist only once, and in one place. This has it's history in the cost and scarcity of computing power and storage resource, and has served us very well for many years now. It is, however, broken. And it's broken because the idea that things can be unambiguously codified consistently by a group of people is wrong.
Anyone who has sent out a wordprocessing or spreadsheet template for people to fill in the blanks will know this, as will anyone who has studied research methods as part of a social sciences degree. (As an aside, I am surprised that I have never read more of the similarities between good user interface design and good research survey design).
Now, with a transactional system we find ways to force unambiguity… Drop down lists and checkboxes (or structured answers on a survey). But the real trouble comes when trying to apply a level of IT order to “unstructured” data (the word processing, spreadsheet, presentation and increasingly image-based information churned out in tremendous volume by modern organisations).
Complex filing systems have generally been the “answer” to an ill defined question of either where to store or where to find things. And unless you have a army of librarians, any structure will not last for long. And that, generally is because of the challenge of trying to get a disparate group of individuals to hold a complex information structure consistently in their own heads.
I've had many colleagues over the years who have pulled their hair in frustration at their inability to get a group of people to come to agreement about some supposedly 'simple' concept like 'product' or 'customer'. Here's an example of how challenging this can be:
Let's take a very simple concept – that of 'food'. And let's try to get to a definition of what is (or is not) food. If you are a Western, non-religious type, then cow (or more specifically beef) is probably food. Unless you are vegetarian, in which case it probably isn't. Pig (pork) might be, unless you're again a veggie, or if you have religion of some description. Horse certainly isn't food, unless you're French (in which case it might be, but then you're almost certainly not vegetarian). The table in front of you right now certainly isn't a foodstuff. Unless it's wooden, and your a woodworm (in which case – congratulate you on mastering the internet – who'd have thought it?).
So let's now take this to the extreme. A substance that is so far away from food that we have primeval gag responses to prevent us eating it. Excrement. Definitely not food.
Unless you're a rabbit. In which case congratulations (although slightly less impressive than surfing woodworm). Or a dung beetle (ditto).
The point of all this? Well, that a concept's meaning is absolutely from the perspective of the person viewing it. And unambiguity of classification is almost impossible to achieve, even for as 'simple' a concept as food.
So, what to do? Well, for structured data systems, forced categorisation remains what is needed. Even if you disagree with what is meant by “credit card number” on a web form, nothing but those magic 16 digits will do. But in the world of documents and images, it has to be said that search rather than classification seems to be working rather well. If you can remember that far back, in the mid 1990s Yahoo! used to be based on a classification system for the web. They've kind of abandoned that approach now…
Why is this important? Well, as organisations start to embrace more collaborative tools for working, a natural inhibitor is going to be perceived requirements to classify and file. These could well kill
initiatives at their inception, and will do so because old world models for managing information that are founded on technological limitations of old just don't work in the modern world.
The key difference between searching and classifying is that in a structured system you theoretically can spot nulls- that is, that you can know if something doesn't exist if you look for it and it isn't there. But the reality is that it might not be there because it's been misfiled, or because you're looking in the wrong place. Now compare that to your experiences of searching the internet: if you know what you are doing, these days you usually will find what you are after at first attempt.
mmitII 