author : David Weinberger

There is a problem with physical information, like books. They have to be stored somewhere, and in one place only (unless you are willing to cover the expense and complexity of duplicates). So, you have to decide on a classification scheme. Which one?

Maybe you go for a subject-based scheme like "Religion, History, Science, ...". But then where do you put the book The Interaction of Religion and Science throughout History?

If you decide to solve the problem by just filing everything alphabetically by author (or even by size, to save space, which is what some libraries do!), then how do you find anything? Well, you need metadata, classification information about the stored information. But there is a problem with physical metadata like catalogues. You can store only a small amount of metadata relative to the data, otherwise the catalogue becomes larger than the library, and it becomes impossible to find anything in the catalogue.

I can't remember the last time I looked at a physical library catalogue. Nowadays, in the world of computers and the Web, we can just type a phrase into a search engine (anything from the library's electronic catalogue, to Google), and out pops the answer. The point that Weinberger eloquently makes is that we are no longer restricted by the old physical limitations: our electronic catalogue can have more, much more, information in it that just a computerised card catalogue.

So, now as in the past, if you know the author (metadata) you can find the book (data). But now, additionally, if you know (a quote from) the book (metadata), you can find the author (data). It's that additional capability that gives us so much more power, and is the subject of this wonderful book.

Weinberger argues that this new power requires a new approach to organising data: don't! Or rather, don't impose a single, up-front, order, but let multiple orders emerge from user interactions (tagging, etc). Prescriptive ontologies are doomed to failure because they are not flexible enough (if we want to do something novel, some new weird and wonderful mash-up, a pre-existing ontology probably doesn't cover it); and they are hard work (someone has to go and tag all the information correctly before it can be used). But user tagging (as with Flikr, iTunes, Amazon, etc) can provide all the benefits with none of the drawbacks (well, modulo manipulative tagging to influence retrieval results).

This leads to:

Include all the data, with all its messiness, ambiguities, contradictions, and so on, because who knows what is going to be useful or relevant. Postpone filtering it until you access it, because you don't know until then what filters you want. The key argument is about messiness: there is no one neat way to organise information, even if we decide on a particular way to organise it:

This is related to Lakoff's work on the way we categorise: we don't think in neat tree-like hierarchies (so much the worse for object orientation), but messy networks of prototypes and family resemblances. And those networks change depending on what we are doing. Not only is it true that if you have a hammer, everything looks like a nail, but rather more significantly, if you need to drive a nail, it's amazing what looks like a hammer. (What, you've never hammered something down with the butt end of a screwdriver?) Affordances depend on what you are trying to do. So much the worse for any form of pre-defined categorisation (yes, Semantic Web, I'm looking at you).

Weinberger goes on to show how we can link the advantages of flexible metadata to physical objects.

But the times they are a-changing. Old fashioned business models, where your metadata needs to be protected because it is part of your business benefit, no longer hold.

Businesses that don't realise will resort to ever more desperate measures to protect "their" metadata, until they are overtaken by those who have worked out a more realistic business model.

This is a super book. Read it to see how the world of information has been radically altered by the move away from the physical, and think how much more change is to come. For example, what changes to education are needed, to move from teaching "how to find something useful" to "how to combine easily found somethings into something else useful"? Vernor Vinge's novel Rainbows End gives some feel for this in a fictional setting: Weinberger shows us how reality is getting there.

Knowledge used to be gatekeepered, curated and preserved in serious books. Now we have the Web, and all its diverse chaotic formats and voices. Overload! Crisis! Catastrophe!

Well, no. In this lovely book, Weinberger argues strongly that our conception of knowledge has been blinkered by those very gatekeepers, curators and formats. Books, which he dubs “long-form thinking”, have their problems: they are one-way (author to reader), and closed format (material chosen and packaged into book-length form, implying that is all there is to be siad on the subject). It is only because we are so used to books that we don’t see their disadvantages and limitations very clearly. (And Weinberger is conscious of the irony of making this argument in book form.)

He examines the current state of this “knowledge messiness”, taking the argument a whole stage further than the information messiness covered in his previous long-form work.

The web allows conversations, hyper-linked non-linear presentation, and an unbounded, “bottomless”, format. It allows diversity: anyone can contribute, not just those previously allowed by the gatekeepers. This openness has its benefits, particularly allowing “networked knowledge”, bringing many eyes, viewpoints and skills to bear. It also has its downsides: it allows fools, and trolls, and confusion, and crud. But then Sturgeon’s law applies to books, too.

Weinberger carefully dissects these issues, showing where, when and how web-based knowledge can be superior to book-based. Books have helped hide the fact that knowledge really is messy, and that even when we have access to it, we will often not come to agreement. The rise of the messy web is not to be lamented, as having destroyed the clean, calm, long-form presentation of books; it should be celebrated for exposing the fundamental messiness of reality, and exploited to help us navigate that messiness. He finishes off with a discussion of how to help ensure the web can provide the best support for messy networked knowledge: open access, metadata, links, and, of course, education.